Arabidopsis seed-specific vacuolar aquaporins are involved in maintaining seed longevity under the control of ABSCISIC ACID INSENSITIVE 3

Highlight Arabidopsis tonoplast intrinsic protein TIP3;1 and TIP3;2 are shown to play a role in seed longevity under the transcriptional control of ABI3.

α-TIP was first purified and identified from Phaseolus vulgaris cotyledons, and the water channel activity of α-TIP is regulated by its phosphorylation (Johnson et al., 1989;Maurel et al., 1995). This protein is specifically expressed during seed maturation and is most abundantly accumulated in dry mature seeds. Immunoelectron microscopy experiments revealed its specific localization to the membranes of seed PSVs in cotyledons and axes (Johnson et al., 1989). α-TIPs are highly conserved proteins that are widely distributed in dicots and monocots (Chaumont et al., 2001;Sakurai et al., 2005;Reuscher et al., 2013;Zhang et al., 2013). Arabidopsis contains two α-TIP orthologues, TIP3;1 (also known as α-TIP) and TIP3;2 (also known as β-TIP). Recently, using fluorescent protein-fused TIP3s, Gattolin et al. demonstrated that TIP3s are dual localized to both the tonoplast and plasma membrane during seed maturation and seed germination (Gattolin et al., 2011).
Single TIP gene loss-of-function mutants do not show obvious phenotypes, probably due to the functional redundancy between different TIPs. The tip1;3/tip5;1 double knockout mutant displays an abnormal rate of barren siliques, indicating that TIPs expressed specifically in pollen contribute to plant reproduction (Wudick et al., 2014). TIP3;1 and TIP3;2 are seed-specific TIP3 isoforms in Arabidopsis and may be the only types of AQPs in mature embryos (Gattolin et al., 2011). Screening of the directly regulated targets of ABSCISIC ACID INSENSITIVE 3 (ABI3), which is a master regulator that controls seed maturation (Parcy et al., 1994;Suzuki and McCarty, 2008), by genome-wide chromatin immunoprecipitation (ChIPchip) revealed 98 Arabidopsis genes including TIP3;1 and TIP3;2 as targets (Monke et al., 2012). The transcription factor ABI3 protein contains a plant-specific DNA-binding domain, designated as the B3 domain (McCarty et al., 1991;Giraudat et al., 1992), and activates numerous seed-specific genes. abi3 mutants exhibit pleiotropic phenotypes to various degrees, depending on the ABI3 alleles, such as desiccation intolerance, decreased seed longevity, abscisic acid (ABA) insensitivity, and lack of chlorophyll degradation (Koornneef et al., 1984;Ooms et al., 1993;Rohde et al., 2000;Tesnier et al., 2002;Delmas et al., 2013). The levels of many seed-specific proteins including late embryo abundant (LEA) proteins, small heat shock proteins (sHSPs), and seed storage proteins are markedly reduced in abi3 seeds (Parcy et al., 1994;Wehmeyer et al., 1996;Wehmeyer and Vierling, 2000). These genes are directly or indirectly regulated by ABI3, and such decreases in protein expression and accumulation may result in the pleiotropic phenotypes of the abi3 mutants (Kotak et al., 2007;Roschzttardtz et al., 2009;Park et al., 2011;Delmas et al., 2013).
Recently it was shown that HvTIP3;1 plays a key role in preventing the coalescence of small PSVs in barley aleurone cells (Lee et al., 2014). However, the function and physiological role of TIP3s in mature seeds are largely unknown. In the current work, the function of TIP3s in seed longevity was investigated. The data suggest that both TIP3 isoforms play roles in maintaining seed longevity and function under the control of ABI3.
Arabidopsis seeds were surface sterilized for 20 min in 10% bleach and washed five times with sterile water. Sterilized seeds were incubated for 48 h at 4 °C in the dark, followed by germination on Murashige and Skoog (MS) medium containing MS salts, 10 g l -1 sucrose, and 8 g l -1 agar, pH 5.8. One-week-old seedlings were transferred to soil and grown in a growth chamber (22 °C, with a 16 h light/8 h dark photoperiod). To collect siliques at different developmental stages, blooming flowers were first marked by tying with cotton threads on the day of pollination. Mature seeds were harvested, dried, and stored at 20 °C.

Plasmid construction and transgenic plants
Mutant TIP3;1 promoters were generated by PCR-directed mutagenesis using a construct containing the 2 kb TIP3;1 promoter as a template (primers are listed in Supplementary Table S1 available at JXB online). WT and mutant promoters were cloned into the pCambia1300 plasmid and transformed into Col. Ten independent T 2 transgenic lines per construct were randomly selected to determine the levels of β-glucuronidase (GUS) expression using real-time PCR (RT-PCR).
For the TIP3;1-RNAi (RNA interference) construct, the PDK intron product amplified from pKannibal was cloned into the PHB vector (Mao et al., 2005) to generate the PHB-RNAi vector. Then, the 2 kb TIP3;1 promoter was cloned into the PHB-RNAi vector by replacing the 2 × 35S promoter to generate the PHB-Pro TIP3;1 -RNAi vector. Then, the TIP3;1 cDNA fragment was amplified and inserted in reverse orientation into both sides of the PDK intron. The PHB-Pro TIP3;1 :TIP3;1 RNAi plasmid was transformed into the tip3;2 mutant background. RNAi transgenic lines (TIP3;1-RNAi/tip3;2) were obtained, and TIP3;1 expression levels in T 3 homozygous transgenic plants were analysed by RT-PCR and western blot analysis. T 4 homozygous lines were used for germination and the controlled deterioration test (CDT). For the Pro 2×35S :TIP3;1 or Pro 2×35S :TIP3;2 construct, TIP3;1 cDNA or TIP3;2 cDNA was cloned into the multiple cloning site of the PHB vector.

RNA extraction and quantitative RT-PCR analyses
Total RNA was isolated from dry mature seeds, siliques, imbibed seeds, and leaves using RNAiso for Polysaccharide-rich Plant Tissue (TaKaRa, Otsu, Shiga, Japan) according to the manufacturer's instructions. Then, 1 μg of total RNA was reverse transcribed using a Primescript RT Reagent Kit with gDNA Eraser (TaKaRa). Quantitative RT-PCR (qRT-PCR) analyses were performed using the SYBR Green method (SYBR premix EX taq, TaKaRa) with the StepOnePlus™ Real-time PCR System (Applied Biosystems). The PCR program was as follows: 95 °C for 10 s, 60 °C for 35 s, repeated for 40 cycles.
To quantify gene expression in seed development and seed germination, geNorm 3.5 software was used to select four reference genes (PP2A, CYP5, ACT7, and ACT8) for seed development and another four reference genes (PP2A, Ef1a, CYP5, and ACT8) for seed germination from nine tested genes. The normalization factor, normalized GOI (gene of interest) quantity, and stand deviation (SD) of the normalized GOI quantity were calculated according to the geNorm manual (Vandesompele et al., 2002).

Protoplast transformation and promoter activation assays
Protoplast isolation and transformation were performed according to the protocol described previoulsy (Yoo et al., 2007;Wu et al., 2009) with minor modifications. Protoplasts were isolated from rosette leaves of 4-week-old Arabidopsis plants using the tape method as described (Wu et al., 2009).
Promoter activation assays were performed with a dual-luciferase reporter assay system (Hellens et al., 2005). ABI3 and FUS3 cDNAs were cloned into the effector plasmid pGreenII 62-SK, and TIP3;1 and TIP3;2 promoter sequences were cloned into the reporter plasmid pGreenII 0800-LUC, respectively. For transfection, a plasmid mixture (12 μg of effector plasmids and 4 μg of dual-luciferase reporter plasmids) was added to 100 μl of protoplasts (~2 × 10 4 cells). The transfected protoplasts were incubated in the dark for >16 h in the presence or absence of 5 μM ABA. Dual-luciferases activity was assayed using Dual Luciferase Assay Reagents (Promega, Madison, WI, USA) according to the manufacturer's instructions and measured with a Varioska Flash spectral scanning multimode reader (Thermo Fisher). For immunoblotting, 50 μg of effector plasmids were added to 500 μl of protoplasts (~10 5 cells) in the presence of 5 μM ABA.

Recombinant protein purification and EMSA
A partial ABI3 fragment (encoding R559-K720) including the B3 domain was cloned into pET28a (Novagen) and transformed into Escherichia coli. The recombinant protein was induced at 16 °C and purified in its native form using Ni-NTA agarose (Qiagen, Venlo, Limburg, The Netherlands) following the manufacturer's protocol.
Elecrophoretic mobility shift assays (EMSAs) were performed using a LightShift Chemiluminescent EMSA Kit (Thermo Fisher). The B3 domain of recombinant ABI3 protein was incubated with biotin-labelled probes containing different RY motif fragments at 20 °C for 30 min in the binding system [1× binding buffer, 2.5% glycerol, 50 ng μl -1 poly(dI-dC), 50 mM KCl, and 0.5 mM EDTA]. DNA-protein complexes were separated by 6% TRIS/ borate/EDTA PAGE and transferred onto a Hybond-N + nylon membrane (GE Healthcare Life Sciences, USA). Biotin-labelled probes were detected by HRP-conjugated streptavidin and visualized with an ECL Detection Kit according to the manufacturer's instructions.
Yeast one-hybrid and DNA-protein interaction ELISA Yeast one-hybrid assays were performed using a Matchmaker One-hybrid system (Clontech, Mountain View, CA, USA). Three tandem copies of the RY2 element of Pro TIP3;1 (GGCACACATGCATGCTTAGT) and three copies of the RY element of Pro TIP3;2 (CTTGGCACACATGCATAGATATAT) were cloned upstream of the HIS3 reporter gene in the pHISi vector, respectively. These reporter constructs, as well as the empty pHISi vector, were linearized with XbaI and integrated into the genome of the YM4271 strain to generate reporter strains. The reporter strains were transformed again with pGAD424-ABI3 (559R-720K) or empty vector pGAD424, respectively. Yeast transformants were grown on synthetic SD-Leu-His medium, and binding activity was monitored on SD-Leu-His medium supplemented with 30 mM or 60 mM 3-amino-1,2,4-triazole (3-AT).
DNA-protein interaction-enzyme-linked immunosorbent assay (DPI-ELISA) was performed as described in Brand et al. (2010). Full-length glutathione S-transferase (GST)-ABI3 protein was produced in the BL21-Codon Plus strain and purified using Glutathione Superflow Resin (Qiagen). An antibody against GST conjugated with HRP was used to detect the bound proteins.

Controlled deterioration test and basal thermotolerance assay
The CDT was performed as described previously with minor modifications (Tesnier et al., 2002;Oge et al., 2008;Chen et al., 2012). Different Arabidopsis seeds for the test were harvested from plants at the same time, dried, and stored under the same conditions (20 °C in a desiccator containing blue self-indicating silica gel) for at least 2 weeks prior to the experiment unless otherwise indicated. The seeds were equilibrated for 3 d at 15 °C and 85% relative humidity (RH). After equilibration, the seeds were stored at 40 °C and 80% RH in a temperature-and humidity-controlled incubator. The temperature and RH in the incubator were corrected and monitored using a thermohygrometer (Testo 608-H1, Germany). The seeds were stored under these conditions for 1-7 d. After storage at high temperature and RH, the seeds were stored at 20 °C and 33% RH for 3 d and dried to 6% moisture content. Seed moisture content was determined by weighing the seeds before and after drying at 105 °C for 24 h.
The basal thermotolerance assay (BTA) was performed as described (Prieto-Dapena et al., 2006). Seeds used for testing were harvested at the same time, dried, and stored under the same conditions for at least 2 weeks prior to the experiment. The seeds were imbibed in Eppendorf tubes and incubated at 50 °C for 1-4 h. After treatment, the seeds were cooled to room temperature and plated on MS medium. For HgCl 2 treatment, imbibed seeds were incubated at 42 °C for various times in the presence of 50 μM HgCl 2 or 1 mM dithiothreitol (DTT), washed five times in sterile water, and then grown on MS medium.
All germination analyses were performed with four replicates using ~100 seeds per replicate. The germination percentage was calculated as the mean ±SD.

Tetrazolium assay and H 2 O 2 staining
The tetrazolium assay was performed as described by Wharton (1955) with minor modifications. Embryos isolated from imbibed seeds after the CDT were soaked in a solution of 1% 2,3,5-triphenyl tetrazolium chloride (Sigma-Aldrich) and incubated at 22 °C for 12 h. Viable embryos stained red, and non-viable or dead embryos remained unstained. Seed viability was evaluated by examining the staining pattern and colour intensity.

Hydrogen peroxide permeability assay in yeast
Yeast strains Δdur3 (MATa, ura3), Δyap1 (MATa, ura3), and Δskn7 (MATa, ura3) were transformed with pYX212 (or derivatives of pYX212 carrying AQP cDNAs). Yeast transformants were inoculated and grown to the stationary phase. The cells were diluted to an OD 600 of 1.0 with SD-Ura liquid medium. A 10 μl aliquot of 0.1 OD cells was spotted onto SD-Ura medium containing the indicated concentration of H 2 O 2 . Photographs were taken 3 d after incubation at 30 °C.
For the fluorescence assays, WT (THY.AP4) cells and pYX212 or AQP transformants were grown to mid logarithmic phase, incubated with 30 μM CM-H 2 DCFDA (Molecular Probes, Life Technologies) for 45 min, washed five times in 20 mM HEPES (pH 7.0), and finally suspended in HEPES buffer at an OD 600 value 1.0. Fluorescence was measured with the Varioska Flash spectral scanning multimode reader at excitation/emission of 492/527 nm and a temperature of 20 °C.
Hypo-osmotic yeast protoplast swelling assay THY.AP4 yeast cells transformed with pYX212 or derivatives of pYX212 carrying AQP cDNAs were grown to mid logarithmic phase, harvested, washed twice in sterile water, and suspended in SCE buffer (1 M sorbitol, 0.1 M sodium citrate, 10 mM EDTA, 0.2 mM β-mercaptoethanol, pH 6.8) containing 200 U ml -1 lyticase (L2524, Sigma-Aldrich) for 2 h at 30 °C. Following centrifugation, protoplasts were washed twice and resuspended in STC buffer (1 M sorbitol, 10 mM TRIS-HCl pH 7.5, 10 mM CaCl 2 ) to the same OD 600 . Protoplasts were diluted to 0.5 M sorbitol with sterile water using a syringe dispenser. The change of OD 600 value was monitored per 0.1 s with the Varioska Flash spectral scanning multimode reader in a flash mode.

TIP3s are specifically expressed during seed maturation
The Arabidopsis genome contains two TIP3 genes, namely TIP3;1 and TIP3;2. Using qRT-PCR, the temporal expression patterns of TIP3 genes were investigated in a precise manner. Transcripts of TIP3;1 and TIP3;2 began to be detectable in siliques at 12 days post-anthesis (DPA) (Fig. 1A). TIP3 transcript levels increased sharply throughout the maturation phase. Immunoblot analysis also indicated that TIP3s began to accumulate at the same time point (Fig. 1C). Since the antibody raised against the C-terminal peptide of TIP3;1 cannot discriminate between TIP3;1 and TP3;2 (Supplementary Fig. S1 at JXB online), the detected signals represented both TIP3 isoforms. In germinating seeds, the levels of TIP3 transcripts decreased to <1% during the first 3 h after germination (Fig. 1B). Interestingly, the protein levels of TIP3s did not decrease significantly within 24 h, but they started to decrease sharply 48 h after germination (Fig. 1D). At the same time, TIP1s (TIP1;1 and TIP1;2) began to be detectable (Fig. 1D). Fluorescent signals were detected in the seeds of Pro TIP3;1 :GFP transgenic plants, but not in WT seeds ( Supplementary Fig. S2). These data suggest that TIP3;1 and TIP3;2 are specifically expressed in seeds, and the TIP3;1 promoter is active in seeds.

TIP3;1 and TIP3;2 are required for seed longevity
To characterize the effects of loss of function of TIP3 genes, two T-DNA insertion mutants were obtained from the ABRC. PCR analysis of genomic DNA from the mutants confirmed the locations of the T-DNA insertions. SALK_053807 has an insertion in the promoter region between the RY2 and RY3 motif of TIP3;1, and SALK_125353c has an insertion in the first intron of TIP3;2 ( Fig. 2A, B). The expression level of TIP3;1 in tip3;1 seeds was reduced to 30% of that in WT seeds (Fig. 2D). TIP3;2 transcripts were not detectable in tip3;2 seeds ( Fig. 2D), demonstrating that the tip3;2 mutant is transcript null.
This observation prompted the study of whether seed longevity was affected by the null function of the TIP3 genes. BTAs, which partially reflect seed longevity and viability, were first performed. As a general inhibitor of AQPs, mercury binds to the cysteine residue near the pore site and inhibits the channel activity of AQP by occluding the pore (Daniels et al., 1996;Maurel and Chrispeels, 2001;Savage and Stroud, 2007). DTT can function as a scavenger to reverse the inhibition effect of mercury (Martre et al., 2001;Vander Willigen et al., 2006). When seeds were incubated for 24 h at 22 °C in the presence of 50 μM HgCl 2 , seed viability was not impaired. However, the seed germination percentage decreased when incubated at 42 °C in the presence of HgCl 2 ( Supplementary  Fig. S4A at JXB online). The effect was partially reversed by adding 1 mM DTT, suggesting that the activities of AQPs in seeds may be involved in seed longevity.
BTAs were then performed with tip3 mutant seeds. Imbibed seeds were directly incubated at 50 °C for various periods of time. After 1 h of treatment, the germination percentage of tip3;1/tip3;2 double mutant seeds declined drastically compared with WT, tip3;1, and tip3;2 seeds (Supplementary Fig.  S4B at JXB online). After 2 h of treatment, the germination percentages of WT, tip3;1, and tip3;2 seeds decreased significantly but were still much higher than that of the double mutant seeds (Supplementary Fig. S4B). Therefore, tip3;1/ tip3;2 double mutant seeds are more sensitive to heat stress than WT and single mutants.
Seed longevity was further estimated using the CDT. The CDT accelerates seed ageing by increasing the temperature of seed storage and the seed moisture content. The germination percentage of all untreated seeds was ~100% at 7 d after germination. When treated at 40 °C with 80% RH for 4 d, WT, tip3;1, and tip3;2 seeds had a germination percentage of ~90%, whereas that of tip3;1/tip3;2 double mutant seeds was only 30% (Fig. 3C, E). Furthermore, the germination percentages of the seeds of three TIP3;1-RNAi/tip3;2 transgenic lines were further reduced to 1-10% (Fig. 3C, E). The seed germination percentage was <50% for tip3;1/tip3;2 and TIP3;1-RNAi/tip3;2 seeds after 3 d of the CDT, whereas for the tip3;1, tip3;2, and WT seeds, the germination percentage was <50% after 5 d of the CDT (Fig. 3C). Unlike the severe alleles of the ABI3 mutant, abi3-1 and abi3-7 seeds are desiccation tolerant but with decreased longevity Fig. 3. Natural and artificial seed ageing assays showing that TIP3 genes are involved in maintaining seed longevity. (A) Germination and growth of seeds from Col, tip3;2, and three lines of TIP3;1-RNAi/tip3;2 stored for 2 weeks or 18 months. The photographs were taken 6 d after seed imbibition. (B) Germination percentages of seeds 6 d after imbibition. (C) Germination percentages of tip3 mutants and TIP3;1-RNAi/tip3;2 transgenic seeds submitted to a CDT for 1-7 d. The germination percentages were counted 7 d after imbibition. (D) Germination percentages of Ler and abi3-1 seeds submitted to a CDT for 1-7 d. The germination percentages were counted 7 d after imbibition. (E) Germination percentages of tip3 mutants and TIP3;1-RNAi/tip3;2 transgenic seeds after a 4 d CDT. Germination percentages were counted at different time points after imbibition. (F) Germination percentages of Ler and abi3-1 seeds after a 4 d CDT. Germination percentages were counted at different time points after imbibition. Values are the means ±SD of four technical replicates with 100 seeds per replicate. In (C-E) seeds used for germination tests were harvested at the same time and stored for 2 weeks prior to the experiment. (Ooms et al., 1993;Bies-Etheve et al., 1999;Tesnier et al., 2002). abi3-1 mutant seeds were also subjected to the CDT. The seed germination percentage decreased to 10% at 2 d after the CDT and to <1% after 3 d (Fig. 3D), which was even lower than that of tip3 knockdown mutant seeds. At 4 d after the CDT, abi3-1 seeds cannot germinate (Fig. 3F). A tetrazolium assay confirmed that the seeds of the tip3;1/ tip3;2 double mutant and three TIP3;1-RNAi/tip3;2 transgenic lines began to lose viability 3 d after the CDT, which was earlier than observed in the WT, tip3;1, and tip3;2 mutants (Fig. 4A). Again, the abi3-1 mutant was the most sensitive to the CDT, as its seeds started to lose viability after 2 d. Taken together, these results show that TIP3s are required for seed longevity.
Knockdown of TIP3;1 and TIP3;2 results in the elevated accumulation of hydrogen peroxide upon CDT One of the most critical factors that influence seed ageing is the accumulation of reactive oxygen species (ROS) in seeds (Bailly, 2004). ROS lead to lipid peroxidation, DNA damage, and inactivation of enzymes. The genes encoding proteins which can scavenge ROS or repair DNA damage and protect proteins were reported to be involved in seed longevity and seed viability (Sattler et al., 2004;Oge et al., 2008;Waterworth et al., 2010;Chen et al., 2012;Verma et al., 2013;Wang et al., 2014). DAB staining showed that the H 2 O 2 contents in seeds increased during the CDT and accumulated to higher levels in tip3;1/tip3;2 seeds than in WT seeds (Fig. 4B).

TIP3;1 and TIP3;2 are activated by ABI3 during seed maturation
The transcription factor ABI3 is involved in seed desiccation tolerance and seed longevity (Ooms et al., 1993;Tesnier et al., 2002). TIP3 genes are seed-specific genes during seed maturation, and tip3 knockdown mutant seeds exhibit a decrease of seed longevity similar to the seed longevity phenotype of abi3-1 and abi3-7. It was hypothesized that TIP3s may maintain seed longevity under the expressional regulation of ABI3. To test whether the seed-specific transcription factors ABI3 or FUS3 are involved in the regulation of TIP3 gene expression in seeds, the presence of TIP3 gene transcripts in the corresponding mutant seeds was investigated. As expected, TIP3;1 and TIP3;2 transcripts were not detectable in abi3-6 mutant seeds (Fig. 5A). The abi3-6 allele contains a deletion in ABI3 which causes a premature stop codon and leads to translation of a short form protein containing only the A1 domain but not the B1, B2, and B3 domains. The expression levels of TIP3 genes in the fus3-3 mutant decreased ~50% compared with those of the WT (Col) (Fig. 5A). The protein levels of TIP3s decreased significantly in the abi3-6 and fus3-3 mutants (Fig. 5B). The expression levels of ABI3 in abi3-6 and fus3-3 mutant seeds were also analysed. In fus3-3 seeds, ABI3 expression decreased to 50% compared with the WT (Supplementary Fig. S5 at JXB online). Therefore, the reduction of TIP3 gene expression in fus3-3 correlates with decreased expression of ABI3 in fus3-3 mutant seeds. the expression of TIP3 genes was also analysed in other alleles of ABI3, namely abi3-1, abi3-4, and abi3-8. The amount of TIP3 transcripts and the abundance of proteins were reduced to varying degrees in these abi3 mutants (Supplementary Fig.  S6). TIP3;1 promoter activity was also reduced in the abi3-6 mutant. No green fluorescent protein (GFP) fluorescence and GFP protein expression could be detected in isolated abi3-6 embryos transformed with Pro TIP3;1 :GFP ( Supplementary  Fig. S7). Taken together, these results suggest that ABI3 is required for TIP3 gene expression and protein accumulation in mature seeds.
Compared with the empty vector, transient expression of ABI3 in the protoplasts slightly increased the activity of the TIP3;1 and TIP3;2 promoters (Fig. 5C). ABA alone could not activate TIP3 promoters, but addition of ABA to ABI3expressing protoplasts caused drastic induction of TIP3;1 and TIP3;2 promoter activity by 279-and 150-fold, respectively, as indicated by the LUC/REN ratio (Fig. 5C). TIP3 proteins accumulated in ABA-treated protoplasts expressing ABI3, but not in the protoplasts expressing FUS3 (Fig. 5E). Consistent with other seed-specific genes (Parcy et al., 1994), ectopic expression of ABI3 also led to the accumulation of TIP3 transcripts as well as TIP3 proteins in vegetative tissues only when treated with ABA (Fig. 5D, F). These results suggest that ABI3 can activate the expression of TIP3 genes in the presence of ABA and is a transcriptional regulator of TIP3 genes.

ABI3 binds to the RY motifs of TIP3 promoters
The B3 domain of ABI3 has DNA binding specificity and recognizes the RY motif (CATGCA) (Suzuki et al., 1997). Promoter sequence analysis showed that TIP3;1 and TIP3;2 promoters possess three and one potential RY motif, respectively (Fig. 6A). To determine whether RY motifs are critical for the activation of TIP3 promoters by ABI3, transient expression assays were performed with the TIP3;1 promoter mutated in RY1, RY2, and RY3 motifs and the TIP3;2 promoter mutated in the RY motif, respectively. Mutations in RY motifs caused reduction of both TIP3 promoter activities (Fig. 6B), suggesting that the RY motifs are required for the promoter activities of TIP3 genes. To determine further which RY motif is important for TIP3;1 promoter activity, promoters containing mutations in the RY motif were fused to the GUS reporter gene and transgenic Arabidopsis plants were generated. GUS expression in transgenic seeds of different lines was determined by qRT-PCR. Mutation of the RY3 or RY1 motif caused a slight reduction in GUS expression, while mutation of the RY2 motif caused a more significant reduction in GUS expression (Fig. 6C). Additional mutations in RY3 or in the RY1 and RY3 motifs did not further reduce the activities of promoters containing a mutation in the RY2 motif (Fig. 6C). This result suggests that the RY2 motif is essential for TIP3;1 expression in seeds.
The EMSA was performed to test whether ABI3 directly binds to the RY2 and RY motifs in the TIP3;1 and TIP3;2 promoters, respectively. The retarded protein-nucleotide complexes were detected in the presence of the B3 domain of ABI3 protein and biotin-labelled RY motifs from TIP3;1 and TIP3;2 promoters. The binding activity increased with increasing concentration of ABI3-B3 protein (Fig. 6D, F). When unlabelled TIP3;1 or TIP3;2 probes were added to the system as competitors, the levels of retarded complexes   (Fig. 6E, G). Additionally, ABI3-B3 did not bind to unlabelled TIP3;1 or TIP3;2 probes harbouring mutations in the RY motifs, and the levels of retarded complex did not decrease. Yeast one-hybrid assays and DPI-ELISA also showed that ABI3 binds RY elements from TIP3 promoters ( Supplementary Fig. S8 at JXB online).

TIP3;2 facilitates both water and hydrogen peroxide diffusion
In addition to transporting water, some TIPs and plasma membrane intrinsic proteins (PIPs) facilitate H 2 O 2 diffusion across the membrane (Bienert et al., 2007Dynowski et al., 2008;Hooijmaijers et al., 2012). In order to understand the connection between TIP3 transport activity and biological function, the water and H 2 O 2 permeability of TIP3s was analysed. To test whether TIP3;1 and TIP3;2 have water channel activity, hypo-osmotic yeast protoplast swelling assays were performed. Hypo-osmotic shock causes water influx and bursting of yeast protoplasts, which could be monitored by a decrease at OD 600 . The yeast protoplasts expressing TIP3;1 or TIP3;2 burst much more quickly than protoplasts transformed with empty plasmid ( Supplementary Fig. S9 at JXB online), suggesting that both TIP3;1 and TIP3;2 have water channel activity.
To determine whether TIP3;1 and/or TIP3;2 is permeable to H 2 O 2 , yeast cells transformed with TIP3 cDNAs were grown on synthetic medium containing different concentrations of H 2 O 2 . TIP1;1 and PIP2;5, which can facilitate H 2 O 2 diffusion across the membrane, were used as positive controls. Three yeast strains differing in H 2 O 2 sensitivity, namely Δdur3, Δyap1, and Δskn7, were used. The result of the growth test showed that the expression of TIP3;2 significantly reduced cell growth and survival on medium containing H 2 O 2 (Fig. 7A). When TIP3;1 was expressed in yeast, the growth of yeast cells was not significantly changed in the presence of H 2 O 2 compared with the negative controls.
TIP3-mediated uptake of H 2 O 2 was further confirmed by using CM-H 2 DCFDA, a dye which was used to measure the ROS level in living cells. Upon exposure to H 2 O 2 , the intracellular level of accumulated ROS was higher in TIP3;2 transformants compared with cells transformed with the empty vector (Fig. 7B). TIPl;1 showed higher H 2 O 2 permeability than TIP3;2 and PIP2;5. TIP3;1 yeast transformants showed almost the same increase in the intracellular level of ROS compared with cells transformed with the empty vector ( Fig. 7B). These results indicate that TIP3;2 but not TIP3;1 can facilitate H 2 O 2 permeation.

TIP3 proteins are involved in maintaining seed longevity and contribute to the ABI3-controlled seed longevity pathway
The AQP family is highly diverse in higher plants and is represented by >30 members in one plant species. Therefore, AQPs belonging to the same group may have functional redundancy (Li et al., 2013). This observation may explain why only a few plant AQP mutants possess clear phenotypes. The protein sequence of TIP3;1 shares 85% identity with that of TIP3;2, suggesting that these two proteins may be functionally redundant. Indeed, only the tip3;1/tip3;2 double mutant (and not the tip3;1 or tip3;2 single mutant) exhibited a significant difference in seed longevity compared with the WT. Seeds of the tip3;1/tip3;2 double mutant were more sensitive to prolonged storage and artificial ageing than the WT (Figs 3, 4). The abundance of TIP3s in seeds appears to be positively correlated with seed longevity, as the TIP3;1-RNAi/tip3;2 lines were more sensitive to the CDT than the tip3;1/tip3;2 mutant, which exhibits higher levels of TIP3;1 (Figs 2E,3,4).
Seed development can be divided into several phases, namely embryogenesis, seed filling, late maturation, and pod abscission. Two important traits of seeds, desiccation tolerance (i.e. 'the ability to survive complete drying and rehydration') and seed longevity (i.e. 'the ability to survive the dry state for prolonged periods of time'), are acquired at seed filling and the later maturation phase, respectively (Verdier et al., 2013). ABI3 has been shown to be involved in both desiccation tolerance and seed longevity. The severe abi3 mutant alleles such as abi3-4 and abi3-6, which have a short form of ABI3 due to a mutation-induced premature stop codon, are intolerant to seed desiccation. Two mutant alleles, abi3-1 and abi3-7, which contain one or two amino acid substitutions in the B2 and B3 domain, are tolerant to seed desiccation but show reduction in seed longevity (Ooms et al., 1993;Nambara et al., 1994;Bies-Etheve et al., 1999;Tesnier et al., 2002). Other weak abi3 alleles were not reported to have a reduction in seed longevity. The different effects may be related to the differences in downstream target genes affected.
A systematic analysis of the Medicago seed development process by transcriptomic and metabolomic profiling (Verdier et al., 2013), as well as studies in Arabidopsis, revealed that LEA proteins are more closely related to acquisition of desiccation tolerance, whereas sHSPs function in desiccation tolerance and longevity (Wehmeyer and Vierling, 2000;Prieto-Dapena et al., 2006). In Arabidopsis, ABI3 directly activates the expression of the transcription factor HSFA9, and HSFA9 activates the expression of sHSPs in seeds (Kotak et al., 2007). Ectopic expression of HaHSFA9 in Arabidopsis leads to the activation of sHSP expression and results in increased seed longevity as well as enhanced desiccation tolerance in seedlings (Prieto-Dapena et al., 2006). Desiccation tolerance and longevity pathways are also connected and share common components. Several lines of evidence suggest that some (but not all) LEA proteins are implicated in seed longevity. In Arabidopsis, a reduction in the levels of three seed-expressed dehydrins results in decreased longevity (Hundertmark et al., 2011). In Medicago, the four most abundant seed LEA proteins are correlated with longevity (Chatelain et al., 2012). These LEA genes are highly abundant in seeds and also regulated by the ABI3 transcription factor. ABI3 therefore functions as a master regulator that regulates the expression of genes in seeds, including LEA and sHSP genes, and controls both the desiccation tolerance and longevity pathways (Prieto-Dapena et al., 2006;Hundertmark et al., 2011).
Here, TIP3;1 and TIP3;2 were identified as members of ABI3 target genes, and TIP3s were added as new components in the seed longevity regulatory network. TIP3 genes are direct targets of ABI3, as demonstrated by ChIP-chip analysis (Monke et al., 2012). Systematic analyses was carried out to demonstrate that TIP3;1 and TIP3;2 are target genes of the ABI3 transcription factor, providing evidence that the B3 domain of ABI3 can bind directly to the RY motifs in the TIP3 promoters; ABI3 is critical for the TIP3 promoter activity in response to ABA and TIP3 gene expression in mature seeds (Figs 5,6). ABI3 therefore plays a critical role in seed longevity through the expressional regulation of TIP3, sHSP, and LEA genes.

TIP3 transport function and seed longevity
Seed longevity is an important genetic trait for preservation of seed viability and seed quality during storage. Orthodox seeds keep their capacity to germinate before and after storage, but gradually lose their viability during storage, which is influenced by genetic factors and environmental factors. Seed storage temperature and seed moisture content are the two most important factors that control seed deterioration and viability loss during storage (Roberts and Ellis, 1989;Bradford et al., 1993;McDonald, 1999). The underlying mechanism of TIP3s in maintaining seed longevity is not clear. TIP3;2 has dual activities on water and H 2 O 2 permeability ( Fig. 7; Supplementary Fig. S9 at JXB online). Some PIPs and TIPs possess the function to facilitate the permeation of H 2 O 2 across membranes, as demonstrated by growth and survival assays with yeast cells expressing AQPs and by H 2 O 2 -detecting fluorescence assays (reviewed by . These identified AQPs are mostly expressed in vegetative tissues, but the biological significance of H 2 O 2 permeation was not addressed. Here it was found that seed-specific TIP3;2 but not TIP3;1 also mediated diffusion of H 2 O 2 across the membrane in the yeast system. ROS are detrimental to seed longevity due to their deteriorative effects on lipids, nucleic acids, and proteins. Vacuoles potentially have a function in ROS detoxification, but the direct evidence for this is still lacking (Mittler et al., 2004;Smirnoff, 2005). TIP3;2 may be involved in H 2 O 2 permeation and detoxification in seeds. However, the tip3;1 or tip3;2 single mutant does not show a significant decrease in seed longevity and only the tip3;1/tip3;2 double mutant is very sensitive to artificial ageing and accumulates a higher amount of H 2 O 2 in the CDT. This result suggests that TIP3;1 and TIP3;2 are functionally redundant in maintenance of seed longevity, and the decrease in seed longevity of the double mutant is not only caused by loss of H 2 O 2 permeability of TIP3;2. The activities of TIP3s in water permeation are important for seed longevity, since the remaining 30% of TIP3;1 expression in the tip3;1/tip3;2 double mutant resulted in higher seed longevity than TIP3;1-RNAi/tip3;2 (Figs 2D, 3C, E). Consistently, a low concentration of HgCl 2 , which is an inhibitor of AQP, also affect the basal thermotolerance of seeds ( Supplementary Fig. S4A).
Water is an essential element during seed desiccation and seed germination, and plays a critical role in the regulation of various seed metabolic processes. Seed moisture content is an important factor for seed deterioration, and the appropriate moisture content can increase seed longevity (Roberts and Ellis, 1989;McDonald, 1999). Lipid auto-oxidation generates various ROS and causes seed deterioration at a moisture content <6%. Above a 14% moisture content, lipid oxidation may again be stimulated by the activity of hydrolytic oxidative enzymes (Labuza et al., 1972;Roberts and Ellis, 1989;McDonald, 1999;Shaban, 2013). Moreover, under a high moisture content, antioxidant enzymes (such as catalase, superoxide dismutase, and glutathione reductase) gradually lose activity and ROS will be accumulated (Bailly et al., 1996;Bailly, 2004). This suggests that changes in water relations from seed development to seed storage and stable water relations in mature seeds may be important for seed longevity. ROS generation may be caused by over-high and over-low seed moisture content. However, the seed moisture content of tip3 double mutants is not significantly changed compared with that of WT mature seeds. Along with the changes in environmental humidity, TIP3s may mediate cell-cell and intracellular water transport and help embryo cells maintain stable water relations in prolonged storage or stressed conditions. Higher H 2 O 2 accumulation in tip3 double mutants might be caused by impaired water transport regulation during seed desiccation and seed storage as a result of loss of water permeability of TIP3s.
In the present study, evidence was provided that seed-specific TIP3;1 and TIP3;2 play a role in maintaining seed longevity during seed ageing. TIP3;2 but not TIP3;1 functions in H 2 O 2 permeation. ABI3 plays the critical role in seed longevity through the expressional regulation of seed-specific gene expression. TIP3s are new members of ABI3 target genes during seed maturation, which work together with sHSPs and LEAs to control seed longevity.