Complex molecular mechanisms underlying seedling salt tolerance in rice revealed by comparative transcriptome and metabolomic profiling

Highlight Comprehensive analyses of phenotypic, metabolic, and transcriptome data from two genotypes with contrasting salt tolerance provided a more complete picture of the molecular mechanisms underlying seedling tolerance in rice.


Introduction
Rice (Oryza sativa L.) is the staple food crop for most Asian people.Grown in diverse ecologies of the world, rice plants suffer many abiotic stresses, of which salt is the most important constraint limiting rice yield in the coastal regions of the tropics and the inland arid/semi-arid areas (Ondrasek et al., 2011).Developing salt-tolerant rice varieties is the most efficient way to reduce the yield loss caused by salt.
Rice is sensitive to salt stress at both the young seedling and reproductive stages (Lutts et al., 1995), and rice genotypes vary considerably in their tolerance to salt, which involves complex physiological mechanisms (Li and Xu, 2007) and is controlled by many quantitative trait loci (QTLs) (Singh et al., 2007;Zang et al., 2008;Thomson et al., 2010;Wang et al., 2013).Several major salt tolerance QTLs with relatively large effects such as qSKC1, qSNC7, and Saltol have been identified (Lin et al., 2004;Thomson et al., 2010).qSKC1 is a sodium transporter controlling K + homeostasis in rice under salt stress (Ren et al., 2005).Unfortunately, few successful cases have been reported in developing new salt-tolerant rice varieties by using molecular breeding approaches based on these identified salt tolerance QTLs.
Historically, plant salt tolerance was found to be associated with the accumulation of specific metabolites in plant tissues such as proline and trehalose (Ashrafa and Foolad, 2007;Li et al., 2011).At the transcriptomic level, plant responses to salt are known to involve large numbers of genes functioning in stress signaling, transcription regulation, ion transport, and biosynthesis of specific metabolites of complex signaling pathways (Kawasaki et al., 2001;Walia et al., 2007;Cotsaftis et al., 2011;Kumar et al., 2013).As an important phytohormone in plant growth and development, abscisic acid (ABA) is known to be involved in plant abiotic stress tolerances (Ye et al., 2012).Exogenous ABA application was able to offset the osmotic and ion stress effects from salt stress in common bean (Khadri et al., 2007), wheat (Gurmani et al., 2013), and rice (Sripinyowanich et al., 2013) by reducing the sodium concentration and improving osmotic adjustment.According to physiological evidence, rice shoots were more stressed than roots under salt (Cotsaftis et al., 2011).At the transcriptomic level, there was limited overlap of stress-responsive genes between shoots and roots of Arabidopsis (Kreps et al., 2002).However, there is little information regarding the relationship between the metabolomic responses and the transcriptomic responses to salt stress in rice, nor about the relationship between shoots and roots of rice, because most reported studies have focused on either shoots or roots of plants.
In this study, the transcriptome and metabolome profiles in the shoots and roots of two related rice genotypes, IR64 and PL177, with contrasting salt tolerance were investigated under salt stress and salt+ABA conditions.Our results revealed organ-specific transcriptome reprogramming and primary metabolite changes in the two genotypes in their response to the treatments, which provides insights into the molecular networks underlying salt tolerance in rice.

Plant materials and stress treatment
A salt-tolerant F 6 pyramiding line, PL177, derived from a cross between two IR64 introgression lines (Supplementary Figs S1 and S2 at JXB online) and its background parent, IR64, were used in this study.IR64 is moderately sensitive to salt stress (Kumari et al., 2009).The seedlings were cultured in Yoshida solution in a box containing 40 plants of each genotype plus 20 plants of the salt-sensitive standard, IR29 (Yoshida et al., 1976), for 11 d after germination.
The salinity treatment was applied by adding NaCl in three steps to reach a final concentration of 140 mM, as described by Walia et al. (2007) with minor modifications (Supplementary Fig. S3 at JXB online).The seedlings were independently subjected to one of three different treatments: no-stress control, salt stress (140 mM NaCl), and salt+ABA (140 mM NaCl+10 μM ABA), with three replications per treatment.The seedlings were grown in a phytotron with 29/22 °C day/night temperatures and a minimum relative humidity of 70%.The pH of the nutrient solution was adjusted daily to 5.5 by adding sulfuric acid and was refreshed every week.

Phenotyping and physiological characterization
The yield performances of the two genotypes under normal and salt stress conditions were evaluated in the summer of 2010.The two genotypes were grown in 2 m 2 plots with three replications for each genotype under normal field conditions in an experimental farm of the Institute of Crop Sciences, Chinese Academy of Agricultural Sciences, Beijing.Under the salt stress conditions, the two genotypes were grown in three concrete ponds at the Tianjing Academy of Agricultural Sciences, which were irrigated weekly with salty water (0.8%) from the seedling stage (7 d after transplanting) until maturity.At maturity, grain yield and related traits were measured from 12 plants of each genotype.
After 11 d under three different treatments, plants of PL177 and IR64 were visually scored for their symptoms of salt injury using the modified standard evaluation system (SES) for rice (Gregorio et al. 1997, Supplementary Table S1 at JXB online).Final scoring and sampling were done when the sensitive standard IR29 scored 7.Each sampled plant was then measured for its shoot and root lengths (cm).The dry weights of the root and shoot samples were measured after being oven dried at 70-75 °C for 72 h.Relative water content (RWC) of each plant was determined as a percentage using the equation: RWC (%)=(FW -DW)/(TW-DW)×100, where FW, DW, and TW are fresh weight, dry weight and total weight of each plant (Larbi and Mekliche, 2004).The sodium and potassium concentrations and sodium/potassium ratio (Na + /K + ) in roots and shoots were measured according to the method described by Zang et al. (2008).

Metabolite profiling and data analysis
Metabolites from the sampled shoots and roots of both genotypes of each replication under the different treatments were extracted and measured based on their chromatograms and mass spectra using the GC-MS Postrun Analysis (Shimadzu) software as described previously (Zhao et al. 2013).Specific mass spectral fragments were detected in defined retention-time windows using the mass spectral libraries of NIST08, NIST08S, and Wiley 9 in the public domain mass spectra library of the Max Planck Institute in Germany (http:// csbdb.mpimp-golm.mpg.de/csbdb/).The quantification of each metabolite was based on its specific peak area.Further confirmation of most identified amino acids, organic acids, and sugars was performed by standard addition experiments using the pure authenticated compounds.In total, 88 primary metabolites were identified in the present study (Supplementary Table S2 at JXB online).

Total RNA isolation and microarray analysis
Total RNA was extracted from liquid-nitrogen-frozen shoot and root samples using TRIzol reagent, and purified and concentrated using an RNeasy MinElute Cleanup kit (Qiagen).Two micrograms of total RNA was used for microarray analysis by CapitalBio Corporation (Beijing) as described previously (Wang et al., 2011).Differentially expressed genes (DEGs) were identified using the empirical criterion of a greater than 2-fold change and a significant q value (false discovery rate-adjusted P value) of <0.05 based on three independent biological replicates.Cluster analyses of the identified DEGs were performed using the complete linkage hierarchical analysis of TIGR MeV 4.2 software (http://www.tm4.org).Gene Ontology (GO) enrichment/over-representation analysis was performed using AgriGO (http://bioinfo.cau.edu.cn/agriGO/;Du et al., 2010).Metabolic pathways analysis was carried out using MapMan software (version 3.5.1R2)(http://mapman.gabipd.org/;Thimm et al., 2004) Real-time PCR confirmation of salt-responsive genes A set of 10 DEGs was selected to confirm the expression level of microarray results using real-time-PCR (RT-PCR).The sequences corresponding to the genes were obtained from the rice genome sequences database (TIGR), and the sequences of exons of each gene were used to design the RT-PCR primers using Primer3 software (http://frodo.wi.mit.edu/); the Actin gene was used as the internal control.The same RNA samples for the microarray analysis were used for RT-PCR, as described previously (Wang et al. 2011).

Data availability
The entire set of original microarray data has been deposited in NCBI's Gene Expression Omnibus under GEO Series number GSE58603.

Results
Phenotypic and genotypic differences in salt tolerance, yield, and related traits under salt and salt+ABA treatments Under normal field conditions, PL177 showed no difference from IR64 for yield and related traits except for having a slightly increased height by 3.6 cm (Fig. 1A, Supplementary Table S3 at JXB online).Under consistent salt stress from Downloaded from https://academic.oup.com/jxb/article-abstract/67/1/405/2885171 by guest on 24 January 2019 the seedling stage to maturity, PL177 was 6.5 cm taller with 1.9 (33.9%) more tillers and 2.8 g (54.9%) higher yield per plant than IR64, indicating that the improved yield of PL177 under salt stress was due primarily to its improved seedling salt tolerance.
Significant differences were detected between IR64 and PL177 for salt damage scores and related traits measured under all three treatments at the seedling stage (Fig. 1B, 1C, Supplementary Table S3).Under the non-stress control and salt+ABA treatments, no significant differences were observed between IR64 and PL177 for all measured traits.However, under salt stress, salt tolerance-related trait values of both genotypes were significantly reduced but to a significantly lesser degree for PL177 than for IR64.Exogenous ABA significantly improved the trait values of IR64 for shoot dry weight (SDW), root dry weight (RDW) and RWC, but not for plant height (PH) and root length (RL).For PL177, exogenous ABA also improved its salt tolerance, SDW, and RWC, but caused no difference in RL and reduced its PH.These results indicated that exogenous ABA could remarkably offset the detrimental effect of salt stress on seedling growth of IR64.
In the absence of NaCl and ABA, IR64 and PL177 showed no differences in Na + and K + concentrations in shoots and roots (Fig. 2, Supplementary Table S4 at JXB online).Salt stress significantly increased Na + concentrations and Na + /K + ratios in shoots and roots of both genotypes.However, the genotypic differences in Na + and K + concentrations under salt were more evident in shoots than in roots, indicating a significant difference between IR64 and PL177 in shoot Na + translocation.Compared with IR64, PL177 showed significantly lower Na + /K + ratios in shoots, attributed largely to more Na + translocation from the roots to the shoots in IR64.Thus, the improved salt tolerance of PL177 was at least partially attributed to its much lower Na + concentration in shoots and thus lower Na + toxicity in its shoots compared with IR64.
The salt+ABA treatment had an overall mitigation effect on salt stress.In comparison with the salt-stressed rice seedlings, the shoot K + content in both genotypes was significantly increased by exogenous ABA.ABA also increased the root Na + concentration but had no effect on the root K + concentration in both genotypes (Fig. 2, Supplementary Table S4).However, under salt stress, exogenous ABA significantly reduced the shoot Na + concentrations but increased root Na + concentrations of both genotypes, although this reduction or increase was more evident in IR64 than in PL177.As a result, the Na + /K + ratio was remarkably reduced in shoots and increased in roots in both genotypes.The translocation of Na + from the roots to the shoots was significantly reduced in both genotypes by exogenous ABA, but the genotypic difference in Na + /K + ratio remained significant although much smaller compared with the salt treatment without ABA.

Genotypic differences in metabolite profiling under salt and salt+ABA
Figure 3 and Supplementary Tables S5 and S6 at JXB online show the general pattern of the relative changes of the 88 measured primary metabolites (19 amino acids, 24 sugars, 29 organic acids, and 16 other small molecular components) in shoots and roots of PL177 and IR64 in response to the salt stress and salt+ABA treatments.In general, the level of virtually all measured primary metabolites except for 5-hydroxytryptamine, trehalose, and allantoin, was significantly higher in shoots than in roots of both genotypes and these changes were highly organ specific.
In shoots, 38 of the 88 measured metabolites showed significant genotypic changes (G) in response to the treatments and explained an average of 6.5% of the total phenotypic variation; the treatments (T) caused significant changes (mostly increases) in most (77/88) metabolites and explained an average 58.6% of the total phenotypic variation; the G×T interaction was significant for 32 of the measured metabolites and accounted for an average of 12.7% of the total variation (Supplementary Table S6A).According to the principal component (PC) analyses, PC1 explained 60.1% of the total measured metabolite variation in the direction of the treatments, while PC2 accounted for 19.4% of the total variation that was attributed partially to the treatment differences and partially to the genotypic differences under salt+ABA.PC3 explained 11.3% of the total variation, which was due primarily to the genotypic differences within different treatments (Supplementary Table S7 at JXB online and Fig. 3A).
Supplementary Table S8 at JXB online shows those metabolites that increased significantly under salt and/or salt+ABA treatment in one or both genotypes; these could be classified into two major types based on their differential responses to the salt and salt+ABA treatments.Type I included 20 metabolites in shoots that increased significantly by more than 2.5-fold in one or both genotypes in response to salt but were insensitive to ABA.Of these, adenosine increased most dramatically by >18-fold and 44-fold in PL177 and IR64, respectively.Type II included 16 metabolites that were significantly responsive to both salt and ABA.Of these, proline, glutamine, and fructose were the most notable.The proline content increased by 66-fold and 25-fold in PL177, and by 204-fold and 769-fold in IR64 under salt and salt+ABA, respectively.The glutamine content increased by 4.7-fold and 6.3-fold in PL177, and by 13.3-fold and 2.6-fold in IR64 under salt and salt+ABA, while the fructose content increased by 15-fold and 22-fold in PL177, and by 28-fold and 515-fold in IR64 under salt and salt+ABA, respectively.In particular, the accumulation in shoots of both genotypes under salt and/or salt+ABA was highest for proline, glutamine, fructose, and fructofuranose, indicating their unique roles in rice salt tolerance.
In roots, 32 of the 88 measured metabolites showed significant genotypic changes in response to the treatments and explained an average of 10.1% of the total phenotypic variation; the treatments caused significant changes in 58 of the 88 measured metabolites and explained an average 38.7% of the total phenotypic variation; the G×T interaction was significant for 37 of the measured metabolites and accounted for an average of 15.6% of the total variation (Supplementary Tables S5 and S6B).Again, the maximum direction (PC1) of the metabolite changes in roots was attributed primarily to the treatments, which explained 35.5% of the total measured metabolite variation, and partially to the genotypic difference within the salt+ABA treatment (Supplementary Table S7, Fig. 3B).PC2 accounted for 26.3% of the total variation and was attributed partially to the treatment differences and partially to the genotypic differences under salt stress.PC3 explained 16.9% of the total variation, which was due primarily to the genotypic differences within different treatments (Supplementary Table S7).
The number of metabolites and the overall levels of salt-or salt+ABA-induced metabolic changes were much less in roots than in shoots of both genotypes.In roots, allantoin showed extremely high accumulation in roots of both genotypes under salt and salt+ABA conditions, with its content increasing by 111-fold and 57.7-fold in PL177 under salt and salt+ABA, respectively, and by 41.8-fold in IR64 under salt+ABA, when compared with the non-stress control (Supplementary Tables S2 and S8).Urea was unique, which decreased slightly in PL177 but increased dramatically by 49 and 30 times in IR64 under salt and salt+ABA.The asparagine contents in PL177 and IR64 increased 7.8 and 2.3 times under salt and by 10.3 and 14.4 times under salt+ABA, respectively.The pentitol contents in PL177 and IR64 increased by 3.2 and 3.9 times under salt, and by 7.4 and 8.2 times under salt+ABA (Supplementary Table S8).
Figure 4 shows the metabolic pathways leading to the most primary metabolites examined in this study, revealing two important results.First, salt-induced metabolic accumulation in the shoots of both genotypes primarily involved a few sugars (glucose, fructose, and fructofuranose) plus proline, all of which were ABA mediated.In contrast, salt-induced metabolic accumulation in the roots of both genotypes involved allantoin, urea, and glutamine, which are products of N metabolism.Trehalose was the only sugar with a relatively high accumulation in the roots of both genotypes.Global transcriptomic responses of PL177 and IR64 to salt and salt+ABA treatments Based on the criteria of a greater than 2-fold change and significance at P<0.05 in t-tests, a total of 3738 genes were found to be differentially regulated in at least one sample of the shoots and roots of PL177 and IR64 under salt and salt+ABA, when compared with their respective controls.Cluster analysis clearly differentiated these DEGs into two major clusters in shoots and roots, subclusters between the treatments, and subclusters between the two genotypes within each treatment (Fig. 5).Using 10 randomly selected genes and 160 comparisons of two organs in two genotypes under two treatments of salt versus the control (10 × 2×2 × 2×2), the quantitative RT-PCR results were in good agreement with those from the microarray experiments (r 2 =0.741;Supplementary Fig. S4 and Table S9 at JXB online).Table 1 summarizes the total number of up-and downregulated genes in roots and shoots of both genotypes under two treatments, revealing two patterns of transcriptomic responses of IR64 and PL177 to salt and salt+ABA.First, more genes were differentially regulated in roots by salt+ABA than by salt alone in both genotypes, but the opposite was true in shoots.Secondly, more genes were differentially regulated, particularly up-regulated, in the salttolerant line PL177 than in its salt-sensitive parent, IR64, under both the treatments.Genotype-and organ-specific gene expression changes in response to salt stress Of the total 1636/1187 and 960/887 DEGs identified in the shoots and roots of PL177/IR64, respectively (Table 1), only 70 and 21 genes were commonly up-and down-regulated in the shoots and roots of both genotypes under salt stress (Fig. 6), indicating that the salt-induced transcriptomic responses of rice to salt stress were largely organ specific (Supplementary Table S10 at JXB online).The majority of DEGs in shoots or roots of both genotypes showed the same expression patterns in response to salt stress, which was expected from the fact that PL177 and IR64 share much of the same genetic background.
As shown in Fig. 5, many DEGs detected under salt were genotype specific, which should have contributed to the phenotypic differences in salt tolerance between PL177 and IR64.Compared with IR64, 287 and 365 genes were exclusively up-and down-regulated in shoots of PL177, while 184 and 47 genes were exclusively up-and down-regulated in the roots of PL177 under salt stress (Fig. 6).The 347 specifically up-regulated genes in the PL177 shoot were involved in eight pathways that are known to be associated with salt tolerance in plants (Supplementary Table S11 at JXB online).MapMan software was used to assign genes to functional categories (Supplementary Fig. S5 at JXB online).These pathways included cytoplasmic vesicle (GO:0031410), vesicle (GO:0031982), cytoplasmic membrane-bounded vesicle (GO:0016023), membrane-bounded vesicle (GO:0031988), hydrolase activity (GO:0016798), polysaccharide metabolism process (GO:0005976), hydrolase activity, hydrolyzing O-glycosyl compounds (GO:0004553), and the extracellular region (GO:0005576).These results suggested that sequestration by vacuoles, detoxification, and cell-wall remodeling are key mechanisms in shoots that contribute to the improved salt tolerance of PL177 (Supplementary Fig. S5A).There were only 36 genes that were specifically up-regulated in PL177 roots, which were involved in four major pathways, including the monocarboxylic acid metabolic process (GO:0032787), oxidoreductase activity (GO:0016491), and mono-oxygenase activity (GO:0004497), implying enhanced oxidation-reduction reactions in the roots of PL177 under salt conditions.Surprisingly, six genes involved in photosynthesis (GO:0015979) were also up-regulated in PL177 roots, suggesting their possible roles other than photosynthesis (Supplementary Fig. S5B).
Large numbers of genes of diverse functions were specifically down-regulated in the PL177 shoots (Supplementary Table S11).The most over-represented GO categories of these included catalytic activity (GO:0003824), kinase activity (GO: 0016301), transferase activity (GO: 0016740), photosynthetic membrane (GO: 0034357), and membrane (GO:0016020), showing that energy metabolism and photosynthesis were evidently repressed in the shoots of PL177.
By mapping the DEGs that were up-regulated exclusively in PL177 on the rice genome, we were able to identify 43 candidate genes that fell into the six genomic regions of PL177 introgressed from its donors (Supplementary Fig. S2 and Supplementary Table S12 at JXB online).Of these, several genes of regulatory function were of great interest, including an ABA-responsive element binding protein (Os06g0211200), a Zn-finger, a RING domain-containing protein (Os04g0105100), Annexin p33 (Os06g0221200), and a receptor kinase-like protein (LOC_Os11g07140).

Identification of ABA-mediated pathways for salt tolerance in rice
Because exogenous ABA was able to ameliorate the salt stress effect on both rice genotypes, direct comparison between the transcriptomic data under salt and salt+ABA conditions allowed us to determine DEGs and their involved pathways contributing to enhanced salt tolerance in rice.Figure 7 shows a four-way Venn diagram of the total 4382 DEGs detected in all shoot and root samples of PL177 and  IR64 under salt+ABA treatment compared with those under only salt stress.Although largely tissue specific, the majority of these DEGs were common in both genotypes, and only 52/160 and 63/70 genes were up-regulated, and 113/114 and 137/121 genes were down-regulated exclusively by exogenous ABA in the shoots and roots of PL177/IR64, respectively.There were 244 and 309 genes that were commonly up-and down-regulated in the shoots and 164 and 641 genes commonly up-and down-regulated in the roots of both genotypes, indicating that they were involved ABA-mediated pathways (Fig. 7).GO analysis indicated that these tissue-specific DEGs were functionally enriched in unique functional categories.The overdominant function categories for the up-regulated genes in the shoots of both genotypes included the lipid metabolic process (GO: 0006629), cellular lipid metabolic process (GO: 0044255), and fatty acid metabolic process (GO: 0006631), while the commonly down-regulated genes in shoots were those involved in the response to stress (GO: 0006950) and stimulus (GO: 0050896), and cytoplasmic vesicle (GO: 0031410), implying that ABA-mediated pathways for salt tolerance include lipid and fatty acid metabolism and cytoplasmic transport in the shoots (Supplementary Fig. S5C).
Table 2 shows the list of 66 DEGs that are known to be functionally involved in ABA biosynthesis/signaling/ response, Na + and K + homeostasis, and salt tolerance in plants.Of these, only three genes (protein phosphatase 2C, inwardly rectifying potassium channel, and heat-shock protein DnaJ) were up-regulated in the shoot of one or both genotypes.Thirteen of these genes were up-regulated in the roots of one or both genotypes but down-regulated in the shoots of one or both genotypes.These included three late embryogenesis abundant (LEA) group family proteins, four dehydrin family proteins, one heat-shock transcription factor and heat-shock protein 70.
There were 27 genes that were down-regulated in the shoot of one or both genotypes.These included 12 heat-shock proteins, five protein phosphatase 2C family proteins, an ABC transporter, Aldehyde oxidase-2, OsNCED3, Na/H antiporter Nhx1, Dehydrin 9 and Dehydrin RAB 16D, and one LEA protein.The remaining 23 genes were all down-regulated in the roots of both genotypes or in PL177, and included an ABC transporter-like protein, Na + /H + exchanging protein-like, Potassium channel SKOR, three heat-shock DnaJ proteins, and 16 tetratricopeptide repeat-containing (TPR)-like domain-containing proteins.We noted that four genes (Aldehyde oxidase-2, OsNCED3 and two ABC transporters) involved in ABA biosynthesis and transport (Hwang et al., 2010;Kuromori et al., 2010, Chen et al., 2014) were repressed or down-regulated in the shoots or roots of both genotypes (Table 2).All the results indicated that ABA biosynthesis and transport were affected in both genotypes under salt+ABA conditions.

Association of gene expression changes with metabolic changes under salt conditions
Table 3 shows the relative expression levels of genes involved in either biosynthesis or degradation process of those metabolites that showed strong salt-induced accumulation in the shoots and/or roots of IR64 and PL177.It appeared that strong salt-induced accumulation of three sugars (glucose, fructose, and fructofuranose) in the shoots of IR64 and PL177 could be partially attributed to reduced degradation and partially to increased biosynthesis, while strong salt-induced accumulation of proline in the shoots of both rice lines was due primarily to greatly reduced degradation.The strong salt-induced accumulation of allantoin and urea in the roots of IR64 and PL177 was inferred primarily due to reduced degradation and partially to reduced transportation (urea), because genes for their biosynthesis showed either reduced (urea) or unchanged (allantoin) expression levels when compared with the non-stress control.

Physiological mechanisms of salt tolerance of PL177
Because of its wide adaptability, high yield, and desirable quality, IR64 has been widely grown in south and southeast Asia for several decades, even though it does not have  good tolerance to salt and drought.Previously, we were able to develop many pyramiding lines with greatly improved salt tolerance including PL177.In this study, we demonstrated that PL177 was very similar to IR64 phenotypically under non-stress conditions but with significantly improved seedling salt tolerance, which was reflected by much less sodium accumulation in both roots and shoots and less shoot damage than IR64, apparently resulting from its lower Na + concentration and higher K + concentration in both tissues and less Na + translocation from the roots to the shoots under salt stress.All these results suggested that the physiological mechanisms of salt tolerance in PL177 are salt exclusion in its roots and salt compartmentation in shoots, which results in its improved capacity to maintain a lower Na + /K + ratio in shoots and a relatively higher growth rate under salt.

Metabolic responses underlying salt tolerance in rice
Consistent with previous findings (Sanchez et al., 2008;El-Samad et al., 2010), our results revealed two important aspects of the metabolic responses of rice to salt stress.First, there were general increases of many primary metabolites in both rice genotypes in response to salt stress, and the salt-and/or salt+ABA-induced metabolic responses were much more pronounced in the shoots than in the roots of both genotypes, which were partially ABA mediated and partially ABA independent.Secondly, strong salt-induced accumulation was observed for only a few primary metabolites, which were highly organ specific, with several sugars (glucose, raffinose, fructose, fructofuranose, and tryptophan) plus proline accumulating strongly in the shoots of both genotypes, while three primary products in the N metabolism (allantoin, urea, and glutamine) accumulated strongly in the roots of both genotypes.Physiologically, many of these metabolites can act as osmolytes in the cytoplasm under salt stress, resulting in increased osmotic potential of the cytoplasm in shoot and/or root cells to maintain appropriate cell water status and protect proteins from salt-induced dissociation, while accumulated organic acids and small molecular components might be involved in the compensation of ionic imbalance and carbon metabolism (López-Bucio et al., 2000;Szabados and Savouré, 2010).We further noted that the greatest genotypic differences in salt-induced accumulation were proline and fructose in shoots and allantoin in roots, in which the proline and fructose levels in shoots of PL177 were 5.2-fold and 8.8-fold higher than those in IR64, while the allantoin level in the roots of PL177 was 42.6-fold higher than in IR64 roots only under salt stress.This strongly suggests that salt-induced accumulation of these metabolites may contribute directly to the improved salt tolerance of PL177.This differentiated accumulation of sugars in shoots and N metabolites in roots of rice under salt stress is energetically more efficient, as the biosynthesis and degradation of sugars occur in shoots, and those metabolites in the N metabolic pathways (allantoin and urea) occur in roots (Fig. 4).
In this respect, our results appear to support the notion that stress-induced accumulation of specific metabolites contributes directly to salt tolerance of rice.It should be pointed out that the primary metabolites measured in this study represented only a portion of the metabolome responsive to salt, and other metabolites may also contribute to the salt tolerance of rice.For example, lipids and fatty acids may also contribute to the salt tolerance of rice, as genes involved in lipid and fatty acid metabolic pathways were significantly up-regulated by salt based on the transcriptome data discussed below.

Transcriptomic responses underlying salt tolerance in rice
Extensive transcriptomic studies have revealed that plants can adapt to different environmental stresses by widely reshaping their transcriptome in an organ-specific manner (Wang et al., 2011;Brenner and Schmülling, 2012;Begara-Morales et al., 2014).Consistent with previous studies, large numbers of organ-specific DEGs were detected in both rice genotypes in their responses to salt and salt+ABA.As expected, most up-regulated genes in the shoots of both genotypes included lipid and fatty acid metabolic processes, while the commonly down-regulated genes in shoots were involved in the stress response and cytoplasmic vesicles.In contrast, DEGs commonly up-regulated in the roots of both genotypes were involved in oxidoreductase activity, cation binding, catalytic activity, metal ion binding, and transferase activity, while the commonly down-regulated genes in roots of both genotypes functioned in photosynthesis, photosynthetic membranes, cellular metabolic processes, cellular biosynthetic processes, translation, transcription regulator activity, and the cell membrane.ABA is known to have massive influences on rice responses to salt at the phenotypic, metabolomics, and transcription levels (Du et al., 2013;Gurmani et al., 2013;Sripinyowanich et al., 2013).In this study, exogenous ABA was able to alleviate the salt stress effect on the salt-sensitive line IR64, indicating that the ABA-mediated system for salt tolerance is at least partially impaired in IR64.Indeed, exogenous ABA greatly up-regulated the genes involved in the lipid and fatty acid metabolisms and cytoplasmic transport in shoots of both genotypes under salt stress, suggesting that enhanced lipid metabolism might also contribute to enhanced salt tolerance of rice.Down-regulated genes specifically in the roots or shoots of both genotypes by ABA included most heatshock proteins, heat-shock transcription factors, TPR proteins, protein phosphatase 2Cs, a Na + /H + antiporter, a Na + / H + exchanging protein, and a potassium channel SKOR that are known to be involved in osmotic stress regulation, Na + and K + homeostasis, and stress resistance in ABA signaling pathways (Table 2) (Qiu et al., 2002;Pilot et al., 2003;Rosado et al., 2006;Schapire et al., 2006;Zou et al., 2009;Fukuda et al., 2010;Bassil et al., 2011;Ye et al., 2012).Furthermore, ABA-mediated repression of genes for degradation of key primary metabolites such as proline and allantoin was responsible for their salt-induced accumulation in rice shoots and/or roots observed in this study.We found that those genes for ABA biosynthesis in roots under salt+ABA treatment were significantly down-regulated and so were for those for ABA transport in the shoots and roots, suggesting that genes involved in endogenous ABA signaling are feedback regulated by exogenous ABA.
The transcriptomic differences between IR64 and PL177 provided us with a unique opportunity to gain insights into key candidate genes and the molecular mechanisms underlying salt tolerance in rice.Many genes specifically up-regulated in PL177 under salt conditions are known to be functionally involved in salt tolerance, including Calcinerin B, Potassium transporter 7, and PDR-like ABC transporter in calcium signaling, potassium influx, and detoxification (Gu et al., 2008;Yang et al., 2009;Nuruzzaman et al., 2014).In particular, genes involved in the monocarboxylic acid metabolic process, and oxidoreductase and monooxygenase activities were specifically up-regulated in PL177 roots, implying enhanced oxidation-reduction reactions in the roots of PL177.Furthermore, genes involved in cytoplasmic vesicles, hydrolase activity, the polysaccharide metabolism process, hydrolyzing O-glycosyl compounds, and the extracellular region, and gene families including glycoside hydrolases and lipid transfer proteins related to cell-wall remodeling (Carvalho and Gomes, 2007;Minic, 2008) were specifically up-regulated in PL177 shoots, suggesting that sequestration by vacuoles, detoxification, and cell-wall remodeling are key mechanisms in shoots contributing to salt tolerance in rice.Notably, some genes related to oxidoreductase activity and photosynthesis were enriched in the roots of PL177.Some early studies reported that genes involved in photosynthesis are normally down-regulated in the roots of rice plants under low phosphorus stress (Li et al., 2010) and low nitrogen stress (Lian et al., 2006), but are significantly induced under dehydration stress (Minh-Thu  , 2013).The biological relevance of salt-induced photosynthesis gene expression in roots needs to be further elucidated.Finally, by mapping DEGs specifically up-regulated in PL177 onto the donor segments in PL177, we were able to shortlist only 43 candidate genes for salt tolerance in PL177 introgressed from the donors, including an ABA-responsive element binding protein, a Zn-finger, a RING domain-containing protein, Annexin p33, and a receptor kinase-like protein.Transformation experiments by genetic complementation or knockout to verify the functionality of these genes on salt tolerance are in progress.

Fig. 1 .
Fig. 1.Morphological and yield trait performance of IR64 and PL177 measured under control, salt, and/or salt+ABA conditions.Results are shown as means±SD.

Fig. 2 .
Fig.2.Shoot and root Na + and K + concentrations and Na + /K + ratio in IR64 and PL177 under control, salt, and salt+ABA treatment conditions.Results are shown as means±SD.

Fig. 3 .
Fig. 3. Results of hierarchical cluster analysis of changed metabolite pools.Hierarchical trees were drawn based on detected changed metabolites in shoots (A) and roots (B) of IR64 and PL177 under salt stress (S), salt stress+ABA (SA) treatments.(This figure is available in color at JXB online.)Downloaded from https://academic.oup.com/jxb/article-abstract/67/1/405/2885171 by guest on 24 January 2019

Fig. 5 .
Fig. 5. Cluster analysis of total DEGs under salt and salt+ABA treatments.(This figure is available in color at JXB online.)

Fig. 6 .
Fig. 6.Four-way Venn diagram indicating the number of salt-up-regulated (left) and -down-regulated (right) genes found exclusively in the roots (R) and shoots (S) of two rice genotypes (salt-tolerant line PL177 and its salt-sensitive parent IR64) in the comparison between salt-stressed and non-stress treatments.(This figure is available in color at JXB online.)

Fig. 7 .
Fig. 7. Four-way Venn diagram indicating the number of up-regulated (left) and down-regulated (right) genes found exclusively in the roots (R) and shoots (S) of two rice genotypes (salt-tolerant line PL177 and its salt-sensitive parent IR64) in the comparison between salt+ABA treatment and salt-stressed conditions.(This figure is available in color at JXB online.) Downloaded from https://academic.oup.com/jxb/article-abstract/67/1/405/2885171 by guest on 24 January 2019

Table 1 .
Summary of DEGs in seedlings of PL177 and IR64 under salt and salt+ABA conditions identified by the microarray experiments

Table 2 .
Relative expression levels (compared with salt stress) of 66 genes in shoots and roots of PL177 and IR64 under the salt+ABA treatment, that are known to be functionally involved in the ABA biosynthesis/signaling, and Na + and K + transport, and salt tolerance in plants

Table 3 .
Coordinated changes in gene expression that led to accumulation of some key primary metabolites in shoots or roots of the salt-sensitive rice line IR64 and its salttolerant progeny PL177 under salt (S), salt+ABA (SA) and control (C) treatments Downloaded from https://academic.oup.com/jxb/article-abstract/67/1/405/2885171 by guest on 24 January 2019 et al.