Rice seeds (O. sativa L. ssp. Indica cv 9311) were surface-sterilized with 75% ethanol for 1–2 min, followed by 50% sodium hypochlorite for 25 min and washed with distilled water at least three times. For rice root measurement, sterilized seeds were germinated on ½12-strength Murashige and Skoog (MS) (Murashige and Skoog, 1962) for 7 days at 28°C under the photoperiod of a 12 h light and 12 h dark photoperiod. Arabidopsis (A. thaliana ecotype Columbia (Col-0) was used as wild-type (WT) in this study. Seeds were sterilized with 75% ethanol and 10% bleach (v/v), followed by rinsing with sterile water for at least three times, and then vernalized for 2 days at 4°C. Plants were grown on ½12 MS medium containing 0.8% (w/v) sucrose and 0.75% (w/v) agar in a growth chamber maintained at 22°C and 60% relative humidity under the photoperiod of a 16 h light and 8 h darkness photoperiod.

The coding region of OsSGL was amplified from total RNA of rice Pei’ ai 64S, using an RT-PCR system (Promega, Madison, WI, USA) according to the manufacturer’s instructions. The verified PCR fragment was digested and ligated into binary expression vector p1300-pJITl63, which was derived from pCAMBIA1300 and pJITl63, carrying the constitutive cauliflower mosaic virus (CaMV) 35S promoter. This yielded overexpression vector p1300-pJITl63-OsSGL. Thus, expression of OsSGL was controlled by the CaMV 35S promoter. All enzymes and reagents used for PCR or restriction digestion were purchased from Takara (Japan). Primer pairs used are listed in Supplementary Table S1.

For Arabidopsis transformation, binary construct p1300-pJITl63-OsSGL was introduced into Agrobacterium tumefaciens strain GV3101. Transformation was performed with the floral-dipping method (Clough and Bent, 1998). Transformants (T₁) were selected on a ½12 MS medium supplemented with 40 mg L^-1 hygromycin B. The T₃ lines displaying 100% hygromycin resistance were considered homozygous and used for further experiments.

For rice transformation, plasmid p1300-pJITl63-OsSGL was introduced into A. tumefaciens strain EHA105, and embryogenic calli induced from rice cultivar 9311 (O. sativa L. ssp. Indica) which is widely used as a male parent in the production of high yield rice in china were used as the transformation receptor. Our method followed that described by Xiao et al. (2008). Hygromycin resistance was used to screen positive transgenic plants. Using primers specific for the hygromycin phosphotransferase gene, we performed PCR to confirm the presence of the T-DNA in the transformants. The expression level of OsSGL in positive transgenic plants was determined by semi-quantitative reverse-transcription PCR (RT-PCR) for Arabidopsis and quantitative real-time PCR (qPCR) for rice. Tubulin and Actin2 were used as the internal control in Arabidopsis and rice, respectively. All primers are listed in Supplementary Table S1. We used T₃ homozygous plants in subsequent experiments.

Total RNAs were extracted with TRIzol reagent (Invitrogen, Burlington, ON, Canada) according to the manufacturer’s instructions. Afterward, 1 μg of DNase-treated RNA was reverse-transcribed using a PrimeScript^TM 1st Strand cDNA synthesis kit (Takara) according to the manufacturer’s protocol. The qPCR analysis was conducted with FastStart Universal SYBR Green Master (Roche), and reactions were performed in an ABI 7900HT (Applied Biosystems) at 95°C for 10 min, followed by 40 cycles at 95°C for 15 s and 58°C for 30 s. The internal controls were Tubulin and Actin2 for Arabidopsis and rice, respectively. Each sample was analyzed in triplicate and relative amounts of mRNA were calculated per the comparative threshold cycle method.

For semi-quantitative RT-PCR, two-step RT-PCR method was chosen. Total RNA extractions and cDNA synthesis were performed as described above. All PCR amplifications were conducted in a total volume of 20 μL under the following conditions: 22–27 cycles of denaturation (94°C, 30 s), annealing (58°C, 35 s), and extension (72°C, 30 s). The primer pairs for semi-quantitative RT-PCR and qPCR are listed in Supplementary Table S1.

To determine the subcellular localization of OsSGL in rice protoplasts, we amplified the full-length open reading frame (ORF) of OsSGL without the termination codon. After verification by sequencing, the PCR fragment was cloned into vector pJITL63-GFP to produce an OsSGL-GFP fusion construct (CaMV35S-OsSGL-GFP), thus allowing the fusion gene to be driven by the CaMV 35S promoter. Construct CaMV35S-GFP was used as the control. Transformation mediated by polyethylene glycol (PEG) was performed as described previously (Zhang et al., 2011). After 6–18 h of incubation in the dark, the subcellular location of the recombinant proteins was observed with a confocal laser scanning microscope (Leica TCS 5 STED CW SP5).

For the analysis of osmotic tolerance in Arabidopsis, sterilized seeds of WT and OsSGL transgenic plants were sown in triplicate on ½12 MS medium containing different concentrations of mannitol (0 or 300 mM). After vernalization for 2 days at 4°C in the darkness, plates were transferred to a growth room (22°C, 16 h light/8 h dark) and kept for another 9 days. At the end of the treatment, survival rates were determined. Plants with green cotyledons were designated as survivors. For analyses of the effects of osmotic stress on root length, following vernalization, seeds of WT and transgenic plants were germinated on ½12 MS medium for 3 days. Plants with identical growth performance were transferred to ½12 MS medium containing 300 mM mannitol, and grown for another 12 days. Plates were placed vertically to facilitate comparison of root growth.

To determine drought tolerance in transgenic rice, T₃ homozygous seeds were used for further analysis. Drought assays were performed in a controlled growth chamber PGC15.5 (Percival, Perry, IA, USA). For osmotic stress at the post-germination stage, sterilized seeds were sown on ½12 MS medium containing 0 or 400 mM mannitol for 11 d. Root and shoot were measured at the end of the treatments. For drought assays at seedling stage, WT and transgenic lines were planted in the same buckets. At the four-leaf-stage, watering was withheld and resumed 5 days later. After 3-week recovery, survival rates were determined.

To determine the root morphological parameters at seedling stage, seeds were germinated on ½12 MS medium and kept growing for 7 days in a growth room. Three replicates were performed. Lengths of primary, adventitious and lateral roots were measured. For the determination of adventitious root, the five longest adventitious roots on each seedling were counted. Similar to measurement of adventitious root length, lateral root length was determined with the 15 longest lateral roots on each primary root. For root comparison at the vegetative and reproductive stages, plants of WT and transgenic lines were planted in the same big bucket. After 6 or 12 weeks, roots were rinsed clean by running water, root lengths and dry weights were measured.

Four-week-old rice seedlings were used for biochemical analysis. Both WT and transgenic plants were transferred from the basal nutrient solution to a nutrient solution supplemented with 20% PEG. After 2 days of treatment, proline and soluble sugar contents in harvested tissue samples were measured according to the sulphosalicylic acid method (Troll and Lindsley, 1955) and the anthrone method (Morris, 1948), respectively. The level of MDA was determined with thiobarbituric acid, as described by Kramer et al. (1991).

RNA samples were sent to the Beijing Genomics Institute (BGI) for further RNA-seq analysis. Briefly, total RNA was isolated with TRIzol reagent from the aerial parts of 15-day-old transgenic and WT 9311 rice plants that treated with 20% PEG (w/v) for 6 h. Material from 10 plants of each genotype was pooled for RNA extraction, and sequencing was performed in an Illumina HiSeq 2000 sequencing platform (Beijing, China). Clean reads were mapped to the Rice Annotation Project Database (RAP-DB) using TopHat Version 2.0.9 (Kim D. et al., 2013). The abundance of mapped reads was normalized to RPKM (reads per kilobase of exon model per million mapped reads) (Mortazavi et al., 2008). The DEGseq package was used for identifying genes differentially expressed between sample pairings (i.e., WT vs. transgenic), and P-values were adjusted according to the Benjamini and Hochberg method. RNA-seq read data in this article has been deposited in the National Center for Biotechnology Information (NCBI) SRA database under the accession number SRP094588.

All statistical analyses were performed with Student’s t-tests using SPSS software version 13.0.

The cDNA of OsSGL is 1062 bp long and contains an ORF of 768 bp flanked by a 5′-untranslated region (105 bp) and a 3′-untranslated region (189 bp). This gene encodes a novel protein of 255 amino acid residues with a molecular mass of 26.73 kDa. Bioinformatics analysis showed that OsSGL has a conserved domain of DUF1645 (Pfam PF07816) from amino acid residues 69-227¹. To investigate the possible evolutionary relationships between OsSGL and other DUF1645 proteins, we constructed a phylogenetic tree via the Neighbor-Joining method, using full-length amino acid sequences (Figure 1). This protein is relatively closely related to its homologs in maize, Triticum aestivum, Setaria italica, and Sorghum bicolor. However, the relationship between OsSGL and other rice DUF1645 proteins is more divergent, suggesting that they play different roles in rice.

To elucidate the physiological and functional relevance of OsSGL, we examined its expression profile in young roots, mature stems, panicles at the heading stage, mature leaf blades, and mature leaf sheaths. Our qPCR results revealed that this gene was constitutively expressed in almost all tissues examined, with expression being greatest in the roots and panicles (Figure 2A). We also performed qPCR analysis using 2-week-old rice seedlings exposed to osmotic, salt, or cold stress and found that expression was highly up-regulated by all three as well as by ABA treatment (Figure 2B).

Subcellular localization of OsSGL was determined using a gene construct containing a DNA fragment that encodes OsSGL-GFP fusion protein driven by the CaMV35S promoter. Transient expression of OsSGL was analyzed by confocal laser scanning microscopy. As depicted in Figure 2C, cells expressing the control GFP gene showed cytoplasmic and nuclear distribution of the GFP signals. In contrast, cells expressing the OsSGL-GFP fusion gene showed GFP signal only in the nucleus, suggesting that OsSGL is a nucleus-localized protein.

To examine the biological function of OsSGL, we generated transgenic rice and Arabidopsis plants with a vector containing its full-length ORF, driven by the CaMV 35S promoter. As we had expected, this gene was variably expressed in the transgenic lines. For further phenotypic analysis, we selected homozygous T₃ transgenic Arabidopsis lines B2 and B4 and rice lines L3 and L5, all of which demonstrated higher levels of expression (Supplementary Figures S1 and S2). Strong induction of OsSGL expression by stresses suggested that the gene might be involved in stress tolerance. No obvious difference was observed in shoots between transgenic rice lines and WT under normal condition, while transgenic rice lines exhibited longer shoots than the WT under osmotic condition at the post-germination stage. In contrast, roots from transgenic lines were longer than the WT 9311 roots under both normal and osmotic conditions (Figures 3A–C). To assess drought tolerance at the seedling stage, we withheld irrigation for a short period before returning to a normal watering regimen. By Day 5 of this simulated-drought treatment, all WT plants were severely affected whereas the transgenic lines showed less leaf rolling and wilting. Following the recovery period, the transgenic plants also grew more vigorously (Figure 3D). After recovery, only 37.9% of the WT plants still contained green tissues while an average of 78.7% of the transgenic rice had survived the drought stress (Figure 3E).

The possible physiological mechanism responsible for improved drought tolerance was investigated by noting changes in stress-relevant parameters, i.e., proline, soluble sugar, and MDA contents. Under normal growing conditions, proline and MDA contents did not differ significantly between WT and OsSGL transgenic rice plants, but the level of soluble sugars was slightly higher in the latter. Upon exposure to osmotic stress, proline and soluble sugar contents were significantly elevated in all genotypes, but those increases were much more evident in the transgenic lines than in the WT (Figures 3F,G). Moreover, the MDA contents detected in transgenic plants were significantly lower than that in the WT (Figure 3H).

Consistent with the results obtained with the transgenic rice, similar osmotic-tolerant phenotypes were observed in transgenic Arabidopsis plants hetero-expressing OsSGL. Under normal growing conditions, cotyledon opening and the rate of greening did not differ significantly among genotypes (data not shown). However, under osmotic conditions, the transgenic plants had higher survival rates and produced longer roots (Figures 4A–D). These results demonstrated that OsSGL plays a positive role in osmotic tolerance.

To gain further insights into the mechanism of the improved drought tolerance in OsSGL-transgenic plants, RNA-seq analysis was performed with RNA samples extracted from 15-day-old seedlings after osmotic stress. Among the 47 genes that were up-regulated in transgenic plants, 33 were drought-inducible, as had been revealed by previous microarray experiments (GEO accession number GSE6901). In all, 41 target genes showed changes in transcript levels of more than three-fold in OsSGL-transgenic plants when compared with the WT (P-value < 0.05). These genes encode both functional and regulatory proteins, including stress response proteins, antioxidants, protein kinases, transcription factors, and enzymes involved in metabolism (Table 1). Some reactive oxygen species (ROS)-responsive genes also showed higher expression in the transgenic line, including hydroxyacid oxidase (LOC_Os08g14860), peroxidase (LOC_Os07g44499, LOC_Os07g44480), peroxidase binding protein (LOC_Os01g73200), proline-rich protein (LOC_Os08g15080), and LEA-related protein (LOC_Os05g01680), suggesting that the improved drought tolerance in transgenic plants partially resulted from ROS pathways modulation. We then selected six stress-responsive genes and compared their expression between transgenic and WT plants grown under osmotic stress conditions. qPCR analysis showed an overall pattern of expression similar to that revealed by RNA-seq analysis, confirming the reliability of RNA-seq analysis (Supplementary Figure S3). These data indicated that overexpression of OsSGL in rice affects the expression of a set of stress-responsive genes, thereby conferring greater drought tolerance in transgenic plants.

Overexpression of OsSGL in rice yielded altered phenotypes in root architecture. At the seedling stage, the roots of transgenic plants were approximately 50% longer than those of the WT. Root fresh weights were also significantly higher for those transgenic plants (Figures 5A,B, and Supplementary Table S2). To test whether the root enhancement was maintained during subsequent plant development stages, transgenic and WT plants were further cultivated in the same bucket. Figure 5C shows that transgenic plants still possessed enlarged root systems. The dry biomasses of the root systems of the two tested transgenic lines were 10.3 and 13.8% higher than that of WT, respectively (Figure 5D). In accordance with seedling and vegetative stages, transgenic plants also had larger root systems at the reproductive stage, displaying the increased root dry weight (Figures 5E,F).

The regulatory pathways for root initiation and growth involve CRL1, CRL4/GNOM1, DR01, CKX4, NAC5, NAC9, and WOX11 (Liu et al., 2005; Zhao et al., 2009; Redillas et al., 2012; Jeong et al., 2013; Uga et al., 2013; Gao et al., 2014). To determine whether OsSGL regulates these genes, we performed qPCR analysis with root samples from transgenic plants and found that CRL1, CRL4, DR01, CKX4, and NAC9 were highly expressed. This implied that OsSGL directly or indirectly influences the expression of those genes. In contrast, transcripts of NAC5 and WOX11 were not clearly affected in the transgenic lines (Figure 6A).

To determine whether OsSGL is also involved in auxin and cytokinin signaling, we analyzed the expression of two auxin transport genes (PIN1 and PIN2), three auxin biosynthesis genes (YUCCA4, YUCCA6, YUCCA9), and three cytokinin-responsive Type-A RR genes (RR1-RR3) in roots from 14-day-old seedlings of WT transgenic rice. While transcript levels of YUCCA4, YUCCA6, and YUCCA9 were elevated in the OsSGL transgenic plants, expression of PIN1, PIN2, RR1, and RR3 was repressed, and that of RR2 was not changed (Figure 6B). These results suggested that OsSGL regulates genes for root development by modulating the signaling pathways of cytokinin and auxin.