Rice (Oryza sativa L.) japonica variety ‘Dong-jin’ was used in this study. Dry rice seeds were washed with 70% ethanol and subsequently with distilled water. They were then sterilized with 0.4% NaClO solution for 30 min and washed extensively with sterilized water until the NaClO solution was washed off. Sterilized seeds were germinated and grown on half-strength Murashige and Skoog (MS) medium containing vitamins (Duchefa Biochemie, Haarlem, The Netherlands), 3% sucrose, and 0.7% phytoagar for 8–10 days. Seedlings were transplanted to soil and grown at 28°C under long-day (16-h light and 8-h dark) conditions in a green house.

Total RNA was extracted from various tissues of wild-type and Ubi:sGFP-OsPUB2, Ubi:sGFP-OsPUB3, Ubi:RNAi-OsPUB2, and Ubi:RNAi-OsPUB3 transgenic rice plants by using Easy Spin Plants Total RNA Extraction kit (iNtRON Biotechnology, South Korea) according to the manufacturer’s protocol. RNA was quantified using a spectrophotometer (NanoDrop1000; Thermo Scientific, USA). Total RNA (2 μg) was used to synthesize cDNA by using TOPscript Reverse Transcriptase (Enzynomics, South Korea) and oligo (dT) primers.

Reverse transcription polymerase chain reaction (RT-PCR) was conducted as described previously (Seo et al., 2012). PCR products were separated on a 2% agarose gel and visualized under UV light. Real-time quantitative RT-PCR (qRT-PCR) was conducted on an IQ5 light cycler (Bio-Rad, USA) in 20 μL reaction mixtures by using SYBR Premix Ex Taq II (TAKARA, Japan). The amplification procedures were as follows: 5 min of denaturation and enzyme activation at 95°C, followed by 50 cycles of 5 s at 95°C, 10 s at 55°C, and 10 s at 72°C. The Actin (Os11g06390) gene was used as an internal control, and DREB1B/CBF1 (Os09g35010; CCAAT-binding factor) was used as a positive control for cold stress treatment. GAD (Os03g13300; glutamate decarboxylase), WRKY77 (Os01g40260), MRP4 (Os01g50100; multidrug resistance protein 4), MYBS3 (Os01g50100), TPP2 (Os10g40550; trehalose-6-phosphate phosphatase 2), and DREB1B/CBF1 were cold stress-induced genes (Jain et al., 2007; Su et al., 2010). Gene-specific primers used for PCR are listed in Supplementary Table S1.

In vitro self-ubiquitination assay was performed according to the established protocol described in a previous study (Bae et al., 2011). The Myc-OsPUB2, Myc-OsPUB3, Myc-OsPUB2C281A, and Myc-OsPUB3C280A fusion genes were cloned to pProEx hta vectors (Invitrogen, USA). Briefly, bacterially expressed Myc-OsPUB2, Myc-OsPUB2C281A, Myc-OsPUB3, and Myc-OsPUB3C280A recombinant fusion proteins (500 ng) were incubated for 2 h in the presence or absence of 100 ng E1 (Arabidopsis UBA1), 100 ng E2 (Arabidopsis UBC8), 10 mM ATP, and 0.1 μg/mL Ub with ubiquitination reaction buffer (50 mM Tris-HCl, pH 7.5, 2.5 mM MgCl₂, and 0.5 mM DTT) at 30°C. The reaction products were subjected to immuno-blotting by using anti-Myc (Applied Biological Materials, Canada) and anti-Ub (Santa Cruz Biotechnology, USA) antibodies.

The 3′ ends of OsPUB2 and OsPUB3 coding regions were tagged with synthetic green fluorescent protein (sGFP) or monomeric red fluorescent protein (mRFP) in-frame and inserted into pEarleyGate (pEG) 100 binary vectors that contains the 35S CaMV promoter. The vector was then transformed into Agrobacterium tumefaciens strain LBA4404 by electroporation. The 35S:OsPUB2-sGFP, 35S:OsPUB2-mRFP, 35S:OsPUB3-sGFP, 35S:OsPUB3-mRFP, 35S:NLS-mRFP, and 35S:AtExo70E2-sGFP constructs were expressed in tobacco (Nicotiana benthamiana) leaves by using Agrobacterium-mediated method as described in a previous study (Kim and Kim, 2013). Two days after infection, protoplasts were extracted from the tobacco leaves, and fluorescent protein signals were visualized by fluorescence microscopy (BX51, Olympus, Japan) as described by Byun et al. (2015). NLS-mRFP and AtExo70E2-sGFP were used as nucleus and EXPO marker proteins, respectively.

Rice protoplasts were obtained from 11-day-old seedlings of Ubi:sGFP-OsPUB2 and Ubi:sGFP-OsPUB3 transgenic rice plants before and after 48 h cold treatment and were used for subcellular localization analysis of OsPUB2 and OsPUB3. Fluorescent signals were visualized by fluorescence microscopy in the presence of 50 μM MG132.

Transgenic rice plants were produced by transforming the pGA2897 binary vector plasmids that contained the maize ubiquitin promoter (Ubi) and OsPUB2/OsPUB3 (Ubi:sGFP-OsPUB2, Ubi:sGFP-OsPUB3, Ubi:RNAi-OsPUB2, and Ubi:RNAi-OsPUB3) into the A. tumefaciens strain LBA4404 via electroporation. Callus was generated by germinating wild-type rice (Oryza sativa L. japonica variety called ‘Dong-Jin’) seeds on the callus induction medium (2 mg L^-1 2,4-D, 0.003% casein hydrolysate, 0.4% CHU stock containing vitamins [Duchefa Biochemie], 0.2% gelite, 0.03% L-proline, and 3% sucrose, pH 5.8) and used for Agrobacterium-mediated rice transformation. Transformed callus was selected on hygromycin B (40 mg L^-1) and carbenicillin (250 mg L^-1) containing medium and then transferred to the regeneration medium (2 mg L^-1 kinetin, 4% MS medium containing B5 vitamins, 1 mg L^-1 NAA, 1.2% phytoagar, 2% sorbitol, and 5% sucrose, pH 5.8). All processes during rice transformation were followed as described previously (Byun and Kim, 2014). Transgenic T0 plants were transplanted to soil and independent T4 overexpressing (lines #1 and #2) and T3 RNAi knock-down (lines #1 and #2) transgenic rice plants were used for phenotypic analysis.

Total genomic DNA was extracted from developing leaves of wild-type and transgenic rice plants by using the CTAB (2% CTAB, 2% PVP-40, 1.4 M NaCl, 100 mM Tris-HCl, pH 8.0, and 20 mM EDTA) method as described by Byun and Kim (2014). Total genomic DNA (10 μg) was digested with BamHI or EcoRI restriction enzyme (Thermo Scientific, USA) and separated by electrophoresis on 0.7% agarose gel. Separated DNA on the gel was transferred to Hybond-N nylon membrane, and the blot was hybridized with 32P-labeled hygromycin B phosphotransferase (Hph) probes under high stringency conditions (65°C) as described by Byun and Kim (2014). The autoradiography signals were visualized using the Bio-Imaging Analyzer (BAS2500; Fuji Film, Japan).

Yeast two-hybrid assays were performed according to the method of Bae and Kim (2013) with modifications. The full-length OsPUB2 and OsPUB3 coding regions were inserted into pGAD T7 and pGBK T7 vectors (Clontech, USA), respectively. These constructs or empty vector were co-transformed into AH109 yeast cells by using the LiAc/single-stranded carrier DNA/PEG method. Transformed yeast cells were serially diluted, plated onto four-minus (-Leu/-Trp/-His/-Ade) medium and grown at 30°C for 2 days (Bae and Kim, 2014).

OsPUB2 and OsPUB3 were cloned into pProEx hta vector and pMAL c2X vector (New England BioLabs, UK) plasmids, respectively. Maltose binding protein (MBP) and (His)₆-OsPUB2, (His)₆-OsPUB3, MBP-OsPUB2, and MBP-OsPUB3 recombinant proteins were expressed in Escherichia coli BL21 (DE3). Expressed proteins were purified by affinity chromatography by using Ni-NTA agarose (Qiagen, Germany) for (His)₆-tagged proteins and MBP Excellose (TAKARA, Japan) for MBP fused proteins, respectively. The recombinant proteins were co-incubated with His resin for 3 h at 4°C in the affinity precipitation (AP) buffer (50 mM Tris-HCl, pH 7.5, 1 mM EDTA, 1 mM MgCl₂, 0.5× proteinase inhibitor cocktail, and 0.3% Triton X-100) and captured proteins were washed five times with AP buffer. The bound proteins were eluted, resolved on 10% SDS-PAGE, transferred to PVDF membrane (Millipore, Germany), and subjected to immuno-blot analysis by using anti-MBP antibody (Applied Biological Materials) or anti-His antibody (Applied Biological Materials).

Cell-free degradation assays were conducted as described by Kim and Kim (2013). Cell-free protein crude extracts were prepared from 8-day-old mock or cold-treated rice seedlings by using protein extraction buffer (50 mM Tris-HCl, pH 7.2, 100 mM NaCl, and protease inhibitor cocktail [Roche, Switzerland]). Arabidopsis RGA, which was tagged with Flag, was cloned into pProEx hta vector. Bacterially expressed MBP-OsPUB2, MBP-OsPUB3, and Flag-RGA recombinant proteins were incubated with cell-free extracts for 2, 4, and 6 h at 30°C. Each sample was harvested, boiled, separated by 10% SDS-PAGE, and analyzed by immunoblotting by using the anti-MBP antibody and anti-Flag antibody (Applied Biological Materials). Rubisco was used as a loading control.

For cold stress treatment, 5-week-old wild-type and transgenic rice plants were transferred to cold room at 4°C for 6 days, after which plants were recovered at 28°C and their growth patterns were monitored as described previously (Byun et al., 2015). Total chlorophyll (chlorophyll a + chlorophyll b) was extracted from wild-type and transgenic leaves before and after cold treatment according to Lichtenthaler (1987) with modifications as described by Min et al. (2016). The amounts of chlorophyll a and chlorophyll b were measured at 664.2 and 648.6 nm, respectively, by ELISA microplate reader (VERSAmax, Molecular Devices, USA).

Electrolyte leakage analysis was conducted using 8-day-old rice seedlings before and after cold stress treatments in different time points (0, 6, and 11 days) at 4°C. Seedlings of wild-type and transgenic plants were soaked in a test tube containing 35 mL of distilled water on an orbital shaker (200 rpm) at 28°C for overnight. The electrolyte conductivity of each sample was determined before and after autoclaving by using conductivity meter (Orion Star A212, Thermo Scientific, USA) by the method of Min et al. (2016).

OsPUB2 and OsPUB3 are homologous genes that encode putative U-box E3 Ub ligases with 75% amino-acid sequence identity in rice (Figure 1A). Both predicted proteins contained a single U-box motif and Armadilo (ARM) repeats in their central and C-terminal regions, respectively. OsPUB2 and OsPUB3 are 71–82% identical to putative U-box proteins from monocot plants, such as millet (Setaria italica), false brome (Brachypodium distachyon), and maize (Zea mays), whereas they share 42% sequence identity with AtPUB17 from dicot Arabidopsis thaliana (Supplementary Figure S1). The U-box and ARM domains are well-conserved in both monocot and dicot U-box proteins.

Reverse transcription polymerase chain reaction analysis showed that transcripts of OsPUB2 and OsPUB3 were detected in all tissues examined, including early seedlings, developing and mature leaves, stems, developing seeds, and panicles, with the expression level of OsPUB3 being higher than that of OsPUB2 (Figure 1B). The transcript level of OsPUB2 was elevated in response to high salt (200 mM NaCl for 3 h), dry (15–45% reduction of fresh weight), and low temperature (4°C for 4–8 h) treatments in 10-day-old rice seedlings (Figure 1C). In contrast, the transcript level of OsPUB3 was not affected by stress treatments. Neither of the genes was induced by ABA (150 μM for 3 h). Real-time qRT-PCR assay indicated that the amount of OsPUB2 mRNA began to increase at 1 h after exposure to cold stress and was continuously elevated up to 35-fold after 48 h (Figure 1D). The induction pattern of OsPUB2 was different from that of DREB1B, a cold stress marker gene, the expression of which was maximum at 12 h after cold treatment. The transcript level of OsPUB3 was reduced at 6 h after cold treatment and restored to normal levels thereafter.

Whether OsPUB2 and OsPUB3 possess E3 Ub ligase enzymatic activity was determined by performing in vitro self-ubiquitination assay. Bacterially expressed Myc-tagged OsPUB2 or OsPUB3 recombinant proteins were incubated at 30°C for 2 h in the presence or absence of E1(Arabidopsis UBA1), E2 (Arabidopsis UBC8), Ub, and ATP. Reaction mixtures were separated by SDS-PAGE and subjected to immuno-blot analysis by using anti-Myc and anti-Ub antibodies. As shown in Figure 2, Myc-OsPUB2 and Myc-OsPUB3 yielded high-molecular-mass ubiquitinated bands detected by both anti-Myc and anti-Ub antibodies. In contrast, the exclusion of E1, E2, ATP, or Ub from the incubation mixture abrogated ubiquitinated smear bands. Myc-OsPUB2^C281A and Myc-OsPUB3C^280A, in which the conserved Cys residue in the U-box motif was replaced by Ala, failed to exhibit the E3 ligase activity even in the presence of all reaction components (Figure 2). Overall, the results presented in Figures 1 and 2 indicated that OsPUB2 and OsPUB3 are homologous U-box E3 Ub ligases in rice.

The subcellular localization of OsPUB2 and OsPUB3 was investigated by conducting an in vivo protein targeting experiment. The 35S:OsPUB2-sGFP or 35S:OsPUB3-sGFP chimeric construct was co-expressed with 35S:NLS-mRFP in tobacco (N. benthamiana) leaves by using Agrobacterium-mediated infiltration method. The NLS-mRFP was used as a nuclear marker protein. Protoplasts were prepared from tobacco leaves, and expressed fusion proteins were visualized by fluorescence microscopy. The results revealed that the fluorescence signal of OsPUB2-sGFP was exhibited as small cytosolic punctate bodies and was also found in the nucleus, where it merged with the NLS-mRFP signal (Figure 3Aa). In the case of OsPUB3-sGFP, the fluorescence signals were only exhibited in the cytosolic punctae. When OsPUB2-mRFP and OsPUB3-sGFP constructs were co-expressed in tobacco leaves, the cytosolic punctate fluorescence signals were closely merged, indicating the co-localization of OsPUB2 and OsPUB3 (Figure 3Ab). These punctate localization patterns of OsPUB2/OsPUB3 were reminiscent of EXPOs (Wang et al., 2010; Ding et al., 2014). Indeed, the localization signals of OsPUB2-mRFP and OsPUB3-mRFP closely overlapped with those of AtExo70E2-sGFP, a marker protein of EXPO, in tobacco leaf protoplasts (Figure 3B). We detected very weak punctate localization signals of both OsPUB2 and OsPUB3 in the protoplasts prepared from Ubi:OsPUB2-sGFP and Ubi:OsPUB3-sGFP transgenic rice plants (Supplementary Figure S3). These localization patterns were unchanged in response to cold treatment (Supplementary Figure S3). Thus, OsPUB2 appeared to be localized to the EXPO-like punctate bodies and nuclei, whereas OsPUB3 was predominantly localized to the EXPO-like structure.

The RING/U-box E3 Ub ligases often form stable dimeric complexes, which is critical for their activity (Nikolay et al., 2004; Berndsen and Wolberger, 2014). To determine whether OsPUB2 and OsPUB3 form dimers, we performed a yeast two-hybrid assay. The full-length OsPUB2 and OsPUB3 coding regions were inserted into pGAD T7 and pGBK T7 vectors, respectively, and co-transformed into yeast cells. Transformed yeast cells were serially diluted and plated onto three-minus (-Leu/-Trp/-Ade) and four-minus (-Leu/-Trp/-His/-Ade) growth media. The results showed that both OsPUB2 and OsPUB3 could interact with themselves and formed homo-dimers in the yeast cells (Figure 4A). Considering the high sequence identity (75%) (Figure 1A; Supplementary Figure S1A) and identical subcellular localization patterns (Figure 3) of OsPUB2 and OsPUB3, we speculated that they formed a hetero-dimeric complex. As expected, OsPUB2 and OsPUB3 bound each other to form hetero-dimers in the yeast cells (Figure 4A).

To further confirm the dimeric-complex formation of OsPUB2 and OsPUB3, we performed in vitro pull-down assay. Bacterially expressed (His)₆-OsPUB2 + MBP-OsPUB2, (His)₆-OsPUB3 + MBP-OsPUB3, and (His)₆-OsPUB2 + MBP-OsPUB3 protein mixtures were co-incubated with a Ni-NTA resin. The bound proteins were eluted from the resin and analyzed by immuno-blotting by using anti-MBP and anti-His antibodies. As indicated in Figure 4B, the MBP-tagged OsPUB2 and OsPUB3 proteins were pull-downed from the Ni-NTA resin by the (His)₆-tagged OsPUB2 and OsPUB3, indicating that OsPUB2 and OsPUB3 formed both homo- and hetero-dimeric complexes in vitro.

We next considered the possibility that the formation of a hetero-dimeric complex of OsPUB2 and OsPUB3 might affect their stability. To test this hypothesis, we performed an in vitro cell-free degradation assay. MBP-OsPUB2 and MBP-OsPUB3 recombinant proteins were incubated for different time periods (0, 2, 4, and 6 h) with the crude protein extracts prepared from 8-day-old rice seedlings grown under normal conditions. The levels of MBP-OsPUB2 and MBP-OsPUB3 rapidly decreased over time (0–6 h) in the cell-free extracts. At 4 h of incubation, 76.8 ± 2.0 and 80.2 ± 2.2% of OsPUB2 and OsPUB3 were degraded, respectively (Figure 5A). After 6 h, only 6.0 ± 0.9 and 6.7 ± 3.0% of OsPUB2 and OsPUB3 proteins were detected, respectively. Under our experimental conditions, the apparent half-lives of MBP-OsPUB2 and MBP-OsPUB3 were approximately 150 min and 160 min, respectively, in the mock-treated cell-free extracts (Figure 5B). However, the degradation of MBP-OsPUB2 and MBP-OsPUB3 appeared to be slower in the cell-free extracts prepared from the cold-treated seedlings than in the mock-treated seedlings: apparent half-lives of MBP-OsPUB2 and MBP-OsPUB3 were about 200 min and 280 min, respectively, in the cold-treated cell-free extracts (Figure 5B). We repeated identical sets of experiments using MBP as a negative control. As shown in Figure 5C, the level of MBP was consistent in mock- and cold-treated crude extracts. In addition, RGA, which was degraded by the 26S proteasome complex (Lee et al., 2010; Kim and Kim, 2013), was degraded in a cold-independent manner (Figure 5D). These results suggested that OsPUB2 and OsPUB3 are more stable in cold-treated cell-free extracts than in mock-treated extracts, with the half-life of OsPUB3 being more apparently increased.

Furthermore, when OsPUB2 and OsPUB3 were mixed together in the cell-free extracts, their degradation was markedly delayed (Figure 5E); thus, their apparent half-lives were increased from 150 min and 140 min to 250 min and 350 min, respectively (Figure 5F). On the other hand, both OsPUB2 and OsPUB3 were rapidly degraded in the presence of equal amount of RGA (Supplementary Figures S2A,B), suggesting that slower degradation of OsPUB2 + OsPUB3 hetero-dimer was not due to the increased amounts of proteins in the reaction mixture. This indicated that a hetero-dimeric form of OsPUB2 and OsPUB3 was more stable than the homo-dimers in a cell-free system. Taken together, these results raised the possibility that the stability of OsPUB2 and OsPUB3 are ameliorated in response to cold stress (Figures 5A,B) and by the formation of a hetero-dimeric complex (Figures 5E,F).

To address the cellular roles of OsPUB2 and OsPUB3, we generated transgenic rice plants (Ubi:sGFP-OsPUB2 and Ubi:sGFP-OsPUB3), in which the sGFP-OsPUB2 and sGFP-OsPUB3 genes were ectopically expressed under the control of the maize ubiquitin promoter (Ubi) (Figure 6A). The qRT-PCR analysis results showed markedly increased amounts of OsPUB2 and OsPUB3 transcripts in two different T4 transgenic lines (#1 and #2) of Ubi:sGFP-OsPUB2 and Ubi:sGFP-OsPUB3 plants, respectively, under normal growth conditions (Figure 6B). The overexpression of sGFP-OsPUB2 and sGFP-OsPUB3 proteins was confirmed by immuno-blot analysis by using anti-GFP antibody (Figure 6C). DNA gel-blot analysis showed that these over-expressing transgenic lines were independent (Supplementary Figure S4). In addition, the independent transgenic rice plants (Ubi:RNAi-OsPUB2 and Ubi:RNAi-OsPUB3), in which OsPUB2 and OsPUB3 were suppressed, were constructed using the RNA interference (RNAi) method (Figure 6A; Supplementary Figure S4). The results of qRT-PCR analysis indicated that the expression of OsPUB2 was reduced approximately by 40% in Ubi:RNAi-OsPUB2 (transgenic lines #1 and #2), whereas that of OsPUB3 was decreased by 50% (line #1) to 70% (line #2) in Ubi:RNAi-OsPUB3 (Figure 6D). The transcript levels of OsPUB2 and OsPUB3 were slightly increased (1.1–1.5 times), in Ubi:RNAi-OsPUB3 and Ubi:RNAi-OsPUB2 plants, respectively (Figure 6D). These overexpressing and RNAi-mediated knock-down transgenic rice plants were used for phenotypic analysis in response to cold stress.

Wild-type, T4 Ubi:sGFP-OsPUB2 (lines #1 and #2), and T4 Ubi:sGFP-OsPUB3 (lines #1 and #2) rice plants were grown at 28°C under long day (16-h light and 8-h dark) conditions for 5 weeks. These plants were then subjected to cold stress by transferring them to a cold room at 4°C. After 6 days of low temperature treatment, plants were retransferred to the growth room at 28°C, and their growth patterns were monitored. As shown in Figure 7, most of the wild-type plants exhibited pale green and yellowish leaves after cold stress. They were unable to grow and eventually died (survival rate = 1.7 ± 1.4–7.4 ± 6.4%). In contrast, OsPUB2- and OsPUB3-overexpressing plants showed markedly increased tolerance to cold temperature compared to the wild-type rice plant. The survival rates of Ubi:sGFP-OsPUB2 was 79.1 ± 5.7% (line #1) and 32.5 ± 6.0% (line #2), whereas those of Ubi:sGFP-OsPUB3 were 60.7 ± 9.7% (line #1) and 50.7 ± 4.4% (line #2) (Figure 7).

In addition, 5-week-old wild-type and OsPUB2/OsPUB3-overexpressing plants exhibited similar leaf chlorophyll content (chlorophyll a + chlorophyll b) under normal condition (Figure 8A). Consistent with their tolerance, OsPUB2- and OsPUB3-overexpressing progeny contained higher amount of chlorophyll compared to that in the wild-type plants in response to cold stress. At 1 month recovery of cold treatment (4°C), the leaf chlorophyll content of wild-type plants was 3.5 ± 1.6 mg/g DW, whereas those of Ubi:sGFP-OsPUB2 and Ubi:sGFP-OsPUB3 was 12.5 ± 2.2–13.5 ± 1.1 mg/g DW and 12.6 ± 0.5–16.2 ± 1.7 mg/g DW, respectively (Figure 8A). As a next experiment, electrolyte leakage from 8-day-old cold-stressed seedlings was determined. Wild-type and OsPUB2/OsPUB3-overexpressing plants were grown at 4°C for 0, 6, and 11 days, and whole seedlings were soaked in distilled water for measuring the rates of electrolyte leakage. The data showed that both Ubi:sGFP-OsPUB2 and Ubi:sGFP-OsPUB3 seedlings showed lower ion leakage (9.7 ± 0.2–10.3 ± 1.0% at 6 days and 17.8 ± 2.1–19.5 ± 1.5% at 11 days) than the wild-type (13.9 ± 0.4% at 6 days and 27.0 ± 1.5% at 11 days) plants in response to low temperature (Figure 8B). Overall, phenotypic analyses in Figures 7 and 8 indicated that OsPUB2- and OsPUB3-overexpressors were more tolerant to severe cold stress than the wild-type plants, suggesting that OsPUB2 and OsPUB3 have positive roles in cold stress response in rice plants.

Next, we examined the phenotype of Ubi:RNAi-OsPUB2 (lines #1 and #2) and Ubi:RNAi-OsPUB3 (lines #1 and #2) knock-down transgenic lines. Unlike the overexpressing lines, RNAi-knock-down plants were very similar to wild-type plants in terms of tolerance to cold temperature (Supplementary Figure S5). The total chlorophyll content and electrolyte leakage rate of RNAi-knock-down plants were also indistinguishable from those of the wild-type rice plants (Supplementary Figure S6). These results led us to hypothesize that the basal levels of OsPUB2 and OsPUB3 might be considerably high, and hence, their partial suppression resulted in undetectable effects on the Ubi:RNAi-OsPUB2 and Ubi:RNAi-OsPUB3 knock-down transgenic lines. Alternatively, the suppression of OsPUB2 could be complemented by OsPUB3 and, conversely, the knock-down of OsPUB3 could be rescued by OsPUB2. This assumption seems to be reasonable, since OsPUB2 and OsPUB3 shared high degree of sequence identity (75%), had a similar structure (Figure 1A), and formed a hetero-dimeric complex (Figure 4).

Because the ectopic expression of OsPUB2 and OsPUB3 conferred increased tolerance to cold stress in rice plants, we next determined whether OsPUB2 or OsPUB3 affected the expression profiles of cold-stress responsive genes. Light-grown, 8-day-old wild-type, Ubi:sGFP-OsPUB2, and Ubi:sGFP-OsPUB3 transgenic seedlings were subjected to cold stress (4°C) for 24 h. Total RNA was isolated from the shoot tissue and analyzed using real-time qRT-PCR by using gene-specific primer sets (Supplementary Table S1). The expressions of GAD (Os03g13300; glutamate decarboxylase), WRKY77 (Os01g40260), and MRP4 (Os01g50100; multidrug resistance protein 4), which are cold stress-inducible genes (Jain et al., 2007; Su et al., 2010), were markedly upregulated in Ubi:sGFP-OsPUB2 and Ubi:sGFP-OsPUB3 transgenic plants than in the wild-type plants under both normal and cold-stressed conditions (Figure 9). Although, TPP2 (Os10g40550; trehalose-6-phosphate phosphatase 2) was not induced by cold treatment in the wild-type rice plant, it was 2–6-fold upregulated in OsPUB2/3-overexpressing transgenic plants relative to wild-type plants after 24 h of cold treatment. The transcript level of MYBS3 (Os01g50100) was similar in wild-type and OsPUB2/OsPUB3-overexpressing plants under normal condition, but was higher in the overexpressors after cold treatment (Figure 9). Expression of DREB1B/CBF1 (CCAAT-binding factor) was higher in Ubi:sGFP-OsPUB2/3 plants in normal condition relative to that of wild-type plant, but it became lower after cold treatment. These results suggested that the cold stress-tolerant phenotypes of Ubi:sGFP-OsPUB2 and Ubi:sGFP-OsPUB3 plants were correlated with increased expression levels of cold stress-related genes before or after low temperature treatment.