Brassica napus L. (cv. HuYou15) seeds were obtained from the Shanghai Academy of Agricultural Sciences (Shanghai, China). The seeds were vernalized on wet filter paper in the dark for 1 month at 4°C. The germinated seedlings were transferred to soil in growth chambers under a 16-h/8-h (day/night) cycle at 22 ± 2°C. Col-0 ecotype Arabidopsis thaliana L., Heynh. was used in our study, and the Arabidopsis seeds were surface sterilized (Lindsey et al., 2017). Plates holding the seeds were then maintained in the dark at 4°C for 3 days to synchronize germination, and the seeds were subsequently planted in MS medium under a 16-h/8-h (day/night) cycle.

BnABI3 cDNA was cloned from B. napus L. (cv. HuYou15). BnABI3 cDNA was sub-cloned into the pHB vector to generate 35S::BnABI3 transgenic lines (Supplementary Figure S1) (Mao et al., 2005; Xu et al., 2016). To generate the proAtABI3::AtABI3 and proAtABI3::BnABI3 transgenic plants, 1.6 kb AtABI3 promoter was fused with the AtABI3 and BnABI3 CDS sequences and sub-cloned into the pCAMBIA1300 vector. Transgenic gs1 mutant expressing proAtABI3::BnABI3 chimeric gene was obtained by hybridization. To generate 4Enhp BnABI3-BnABI3GR, the CaMV 35S enhancer tetrad was amplified using pSKI015 as template and cloned into pQDL4R1 to generate pQDL4R1-4Enh. The GR domain was cloned from pTA7002 and the coding region of BnABI3 was cloned from cDNA. Both fragments were fused together and cloned into pQDR2L3 vector to generate pQDR2L3-BnABI3GR. All the reagents and digest enzymes used for PCR amplification and molecular clone were ordered from Takara, Japan. The constructs were transformed into Agrobacterium tumefaciens GV3101 and transformed into Col-0 Arabidopsis by using the floral-dip method (Clough and Bent, 1998). The positive transgenic plants were screened by 50 mg/L Hygromycin (Roche, United States). Real time PCR were performed to detect the gene expression levels in transgenic plants, and the primers used are shown in Supplementary Table S1.

Total RNA was isolated by using the RNeasy Plant Mini Kit (Qiagen, United States) and stored at -80°C. 1 μg total RNA was used for cDNA synthesis using the Transcript First-Strand cDNA Synthesis Mix (TianGen Bio-tech, China). According to the manufacturer’s instructions, qRT-PCR was performed in 96-well plates with 1 μl of a 1:5 dilution of the first-strand cDNA reaction and in a 10 μl volume on a Roche LightCycler^® 96 Sequence Detection System. Transcript levels were measured based on SYBR Green technology (Applied Biosystems Applera, Darmstadt, Germany). Data were analyzed by using LightCycler 96 analysis software 1.1 (ΔΔC_T method). AtACT2 and AtUBQ10 genes were used as internal controls (Xu and Cai, 2017). RT-qPCR primers are shown in Supplementary Table S2.

The 5-week-old BnABI3GR transgenic plants siliques were treated with 20 μM DEX, 20 μM DEX plus 100 μM CHX, or DMSO (mock). The tissues were collected at the indicated time after treatments and put into liquid nitrogen immediately.

For each treatment, 4-week-old plants were transferred to Hoagland’s solution with 100 mM Mannitol, 100 mM NaCl, and 5 μM ABA for 48 h. The leaves sampled at each time point were obtained and stored at -80°C. For the freezing assay, plants were grown in soil in growth chambers at 21°C under long-day (LD) conditions for 3 weeks before the treatments were performed. For cold acclimation experiments, 3-week-old plants were acclimated for 3 days at 4°C in the light, then for 1 h at 0°C, before the temperature were decreased by 2°C per hour. The final sub-zero temperature was maintained for the indicated period before the temperature was again increased at the same rate to 4°C. The plants were then kept at 4°C for 1 day before they were returned to 21°C. Survival rate was counted 2 weeks later.

Quantitative protein analyses were based on 40 seeds each. Total seed proteins, insoluble and buffer-soluble proteins were determined and separated according to a previous method (Bates et al., 2001). Different proteins were quantified using the Bio-Rad (Bio-Rad Laboratories, Hercules, CA, United States) Dc protein assay with three replicates.

Mature seeds (50 mg) were analyzed for crude lipid content, and the method was carried out as described (Bates et al., 2001) with three replicates.

Mature dry seeds of Col-0, abi3-6, gs1 and transgenic plants were fixed in PBS buffer containing 10% formaldehyde, 5% acetic acid, and 50% ethanol and subsequently dehydrated in a (10, 30, 50, 70, 80, 90,100% 1 h each step) graded ethanol series overnight. Seeds were then infiltrated with graded concentrations of LR White acrylic resin (Sigma, Co., St. Louis, MO, United States) at 25, 50, and 75%, each for a minimum of 3 h followed by two changes of 16 h 100% resin. Seeds were placed in fresh 100% LR White acrylic resin in an embedding capsule and polymerized at 60°C for 24 h, then observed under a light microscope (Leica, Germany).

The seeds were sputter-coated with gold and further visualized using a Hitachi JEOL JSM-6360LV SEM (scanning electron microscope).

Tetrazolium red tests were used (Debeaujon and Koornneef, 2000; Debeaujon et al., 2000). Dried seeds of Arabidopsis were incubated in 1% (w/v) tetrazolium red (2, 3, 5-triphenyltetrazolium chloride, Sigma-Aldrich, United States) in the dark at 30°C for 24 h. The seeds were imaged by using a stereomicroscope.

About 1.7-kb long promoter sequences of AtMUM1, AtGATL5, and AtSGR1/2 were amplified from Arabidopsis genomic DNA and inserted into the plasmid reporter vector pGreen II 0800-LUC. The coding sequence of BnABI3 was amplified via PCR and then inserted into the effector plasmid pGreen II 62-SK. Arabidopsis protoplasts were prepared and cotransfected with the constructs in accordance with the instructions for the Dual-Luciferase Reporter Assay system (Promega). The ratio of LUC to REN was determined for the dual-luciferase reporter system (Promega) via a Glomax 20/20 luminometer (Promega) after culturing the protoplasts under low-light conditions for 16 h. The LUN/REN ratio indicates the transcriptional activity.

ChIP and subsequent qPCR (input DNA dilution of 1000×) were performed as previously described (Lavenus et al., 2015). The primers designed to amplify 110–180 bp fragments are listed in Supplementary Table S3. The relative enrichment of the target region was normalized to that of AtACT2. The relative enrichment of BnABI3-HA proteins was analyzed at the regions of target genes promoter. The 35S::BnABI3-HA transgenic line were analyzed via ChIP assays with anti-HA antibodies. The values were normalized to those of the internal controls.

Statistical analysis was undertaken using the Mann–Whitney or Kruskal–Walls with GraphPad Prism 6.0 (GraphPad Software, San Diego, CA, United States). The differences were considered significant when p < 0.05. Multiple alignments and phylogenetic analyses were performed using DNA MAN 6.0 and MEGA 4.1 software (Hall, 2013).

Wild-type (WT) Arabidopsis seeds cannot be germinated on agar-solidified medium containing 10^-4 M uniconazole (Nambara et al., 1991). We found the Arabidopsis line green seed 1 (gs1) from a population of approximately 9000 mutagenized M2 seeds, which germinated normally in media with 2 M uniconazole. In addition to the ability to germinate on 2 M uniconazole, the gs1 mutant was segregated by the production of dark green shriveled seeds in the M3 generation, which could not be germinated after desiccation (Figure 1A). These seedlings produced green seeds that did not germinate when fully desiccated. The experiment of genetic segregation of the 107 M3 seeds showed that 28 seeds were green. This proportion fits the 3:1 hypothesis, suggesting that a recessive single gene mutation contributes to the green seed color phenotype. Reciprocal backcrosses were performed between the WT and gs1 plants, and the data were consistent with the assumption that the green seed phenotype was the result of a monogenic recessive mutation. The genetic association between the green seed phenotype and the ability to germinate on uniconazole was assessed. All the green seeds removed from immature siliques could germinate on uniconazole, indicating that the green seed phenotype and the ability to germinate on uniconazole are due to the same gene mutation (Figure 1C).

The phenotype of gs1 mutant seed is resulted from a single gene mutation and this abnormal seed phenotype is quite similar to the abi3 mutant (Koornneef et al., 1989). Therefore, to identify whether the GS1 mutation is a new allele of the abi3 mutant, we sequenced the entire ABI3 gene and found that a 5 bp deletion in the first exon of AtABI3 which results in a truncated 385 aa protein (Figure 1B and Supplementary Figure S2). Furthermore, we used the ABI3 native promoter fused with the ABI3 CDS sequence to complement the gs1 mutant. Many transgenic gs1 and proAtABI3::BnABI3 lines were generated. After confirmation via qPCR, we selected the lines for further assays (Supplementary Figure S3). The results showed that all transgenic lines have a brown seed coat and acquired desiccation tolerance. Thus, our results demonstrated that GS1 is a new allele of ABI3.

Employing primers generated using GenBank Accession No. LOC106361788 as a reference, we cloned a 2157 bp cDNA encoding a 718 aa polypeptide from the cDNA of B. napus L. (cv. HuYou15) leaves and identified a candidate BnABI3 gene after DNA sequencing. Homology analysis indicated that BnABI3 was 65% identical to A. thaliana AtABI3 (Supplementary Figure S4). The phylogenetic tree was used to examine the evolutionary relationship between BnABI3 and ABI3 proteins derived from other species. These results showed that the ABI3 sequences of higher plants could be classified into two groups. The first group was composed of sequences from monocotyledonous plants, such as Triticum aestivum and rice, while the other group was composed of sequences from dicotyledonous plants, such as A. thaliana, B. napus, Phaseolus vulgaris, and Ribes rubrum. These results show that BnABI3 is more closely related to the ABI3 genes of dicots than that of monocots (Figure 2A).

BnABI3 gene expression levels in different plant stages and tissues, such as seedlings and flowers, leaves, stems, siliques, and seeds at different development stages, were assayed using qRT-PCR. The results showed relatively high expression levels of BnABI3 in seeds and siliques, whereas lower expression levels were observed in stems (Figure 2B). To determine whether the expression of BnABI3 is controlled by environmental cues, the mRNA levels of BnABI3 were measured in the leaves of 3-week-old B. napus plants subjected to various abiotic stresses for a certain period. In response to NaCl treatment, early induction of BnABI3 was observed at 4 h, with greater increases being detected at 12 h (Figure 2C). BnABI3 gene expression gradually increased at 4 h after mannitol treatment, reaching the highest expression level at 24 h (Figure 2D). Maximum induction of BnABI3 was observed after cold treatment for 24 h, with a reduction being detected at 48 h in leaves (Figure 2E). Taken together, these results showed that BnABI3 responds to various environmental cues.

Because B. napus is very difficult to transform, to identify whether BnABI3 could complement the gs1 mutant, we constructed a series of Arabidopsis transgenic lines carrying a proAtABI3:: BnABI3 gene sequence in a Col-0 background and then crossed these lines with gs1 mutant plants (Supplementary Figure S1). To fix the transgene in a homozygous gs1, brown seeds were selected from the basta-resistant plants showing about a 3:1 segregation of seed color. The selected F3 plants were expected to be homozygous for gs1 and hemizygous for the transgene. From this assay, we concluded that the exogenous proAtABI3::BnABI3 transgene was able to complement the gs1 mutant seed coat phenotype (Figure 3A). We also observed protein and lipid content in the BnABI3 transgenic seeds. Compared with gs1 and abi3-6 mutants, there was an obvious difference in the total protein content of the BnABI3 transgenic seeds. Transgenic seeds exhibited the full complement of storage proteins expressed at Col-0 levels (Figure 3B). Moreover, there was a reduced lipid content in abi3-6 and gs1 seeds, and BnABI3 gene expression can absolutely restore the lipid content of abi3 mutants to Col-0 levels (Figure 3C).

The abi3-6 plants are completely insensitive to ABA treatment at the germination stage (Giraudat et al., 1992). Hence, we wanted to determine whether proAtABI3::BnABI3 lines can restore ABA sensitivity. The exogenous ABA biological effects on seed germination and plant growth were further assayed. After 4 days of moist chilling treatment, all seeds showed almost an equal ability to germinate on the media without ABA. As expected, the germination of the Col-0 seeds was inhibited in a gradient of ABA-containing media, but the gs1 mutant seeds displayed no sensitivity up to 10 mM ABA. The dose-response curves of transgene seeds on ABA-containing media were more similar to the Col-0, suggesting that transgene lines partly restored the gs1 mutant ABA sensitivity (Figure 3D). The results also showed that seeds of the gs1 mutant were highly insensitive to ABA and exhibited almost 100% germination. The seeds of WT plants were the most sensitive to ABA, and their germination was increasingly inhibited by 1–10 mM ABA. At ABA concentrations greater than 2 mM, most of the Col-0 plants could not expand their cotyledons and develop into normal seedlings. Interestingly, the seeds of all three transgenic plants showed ABA sensitivities that were intermediate to those of the abi3 mutants (gs1) and Col-0 (Figure 3E,F). The somewhat different sensitivity to ABA during post-germinative growth may show the inability of the BnABI3 protein to completely restore ABA sensitivity in the gs1 mutant background.

Research in the desiccation tolerance of plant seeds has immense basic and applied value (Angelovici et al., 2010). Revealing the responses and adaptations of seeds to rigorous desiccation is essential to improve commercial crop plants by developing these characteristics (Joet et al., 2013). Although ABI3 has important roles in seed desiccation tolerance (Zeng et al., 2003), the underlying molecular mechanism is not well-understood. Previous studies have shown that ABI3 regulates the accumulation of storage proteins involved in desiccation tolerance (Nambara et al., 1992; Khandelwal et al., 2010). However, in the present study, we showed that BnABI3 can regulate seed coat development, thereby playing a role in seed desiccation tolerance. As shown in Figure 4A, we used chloral hydrate to achieve transparency in the premature seeds at 16 days after flowering (DAF) and found that transparency was easily achieved in the gs1 and abi3-6 mutant seed coats at this stage after 3 min, and the seed coats were fragile after treatment (could be broken with moderate pressure). In addition, after storage at 37°C for 24 h, the mature WT seeds appeared normal, but the gs1 and abi3-6 mutant seeds appeared hollow (Figure 4B). Furthermore, we monitored mucilage secretion and found that the gs1 and abi3-6 mutants were defective in this process (Figure 4C). However, this phenotype was not found in the Col-0 seeds, and gs1 and proAtABI3::BnABI3 transgenic plants could fully complement the mutant phenotype in the gs1 background. Thus, the ABI3-mediated intact seed coat could at least partially contribute to seed desiccation tolerance.

During differentiation, the epidermal cells of Arabidopsis seed coat produce and secrete large quantities of mucilage, making it a valuable model for studying cell wall biology. Mucilage is composed mainly of pectin, and also contains cellulose, hemicellulose, etc. By using RT-qPCR analyses, we checked the expression levels of mucilage secretion related genes such as MUM1, MUM2, MUM4, PMEI6, PMEI14, GATL5, and SBT1.7 in Col-0, abi3 mutants and proAtABI3::BnABI3 transgenic plants (Shi et al., 2018) (Figure 4D). We found that the expression levels of MUM1 and GATL5 were altered obviously. The mucilage of the mum1 mutant contains rhamnogalacturonan I that is more highly branched and lacks the ability to be extruded when exposed to water (Arsovski et al., 2009; Francoz et al., 2015). GATL5 is involved in mucilage synthesis and may function to regulate the size of the polysaccharide (Kong et al., 2013).

To investigate the mechanisms underlying the function of BnABI3 in mucilage secretion biosynthesis, we wanted to determine whether MUM1 and GATL5 were immediately downstream of BnABI3 since they were induced in BnABI3GR transgenic plants after dexamethasone (DEX) induction in the presence of cycloheximide (CHX) (Figure 5A–F). The DEX induction assays suggested that MUM1 and/or GATL5 were likely to be downstream targets of BnABI3, which contributes to altered mucilage biosynthesis. Furthermore, we analyzed the promoter sequences of both MUM1 and GATL5 genes. The CATGCA core sequence was reported to be crucial for ABI3 transcription factor binding. There are two and one ABI3 binding motifs in the MUM1 and GATL5 promoter sequences (Figure 5G). We used a LUC-based transactivation experiment. We co-expressed BnABI3 with both MUM1 and GATL5 promoter-LUC fusion reporters in mesophyll protoplasts of Arabidopsis, which resulted in an increase in luciferase activity by 2.4- and 2.7-fold (Figure 5H,I). To provide further evidence for the interaction, ChIP assays were conducted to identify whether the BnABI3 protein could bind to the gene promoters. We used 35S::BnABI3-HA transgenic plants, which contained an HA-coding sequence fused in-frame to the 3′ end of the BnABI3 gene. ChIP-qPCR assays using an anti-HA antibody showed that BnABI3 could bind to the conserved sequence motifs in the MUM1 and GATL5 gene promoters (Figure 5J). These results demonstrate that BnABI3 could target the MUM1 and GATL5 gene promoters and induce their transcription.

Frost is one of the major factors that can cause the fixation of a green color in mature seeds. Exposure of plants from 0 to 1°C sub-lethal frost can fix a green color in seeds, which contributes to major seed industry losses (Johnson-flanagan et al., 1994; Jalink et al., 1998). Because ABI3 controls embryo degreening through Mendel’s I locus (Delmas et al., 2013), we hypothesized that if freezing stress resulted in the down-regulation of chlorophyll degradation genes, BnABI3 overexpression should help to overcome this problem. To verify this hypothesis, we investigated maturing of Arabidopsis pods at 11–13 DAF and found that freezing temperatures resulted in mature green seeds (Figure 6A). However, when 35S::BnABI3 transgenic lines were subjected to similar treatment, the transgenic seeds proceeded to de-green and produced mature brown seeds (Figure 6A). qRT-PCR analyses showed that BnABI3 over-expressing lines expressed much higher levels of SGR1 and SGR2 compared with Col-0 after freezing treatment (Figure 6B,C).

Furthermore, we wanted to determine whether the expression levels of the altered genes in 35S::BnABI3 plants were immediate downstream targets of BnABI3. DEX induction coupled with CHX treatment assays showed that AtSGR1/2 was immediately up-regulated after DEX induction in the BnABI3GR transgenic plants (Figure 7A–D). Furthermore, both promoters possessed the canonical B3 domain binding RY motif CATGCA with variable flanking nucleotides (Figure 7E). We co-expressed BnABI3 with AtSGR1/2 promoter-LUC fusion reporters in mesophyll protoplasts of Arabidopsis, which resulted in an increase in luciferase activity by 3.1 and 3.4-folds (Figure 7F,G). ChIP assays were conducted using 35S::BnABI3-HA transgenic plants to identify whether the BnABI3 protein binds to the AtSGR1/2 promoters. ChIP-qPCR assays using an anti-HA antibody indicated that BnABI3 could bind to the selected regions in the AtSGR1/2 promoters (Figure 7H). Taken together, we concluded that positive regulation of the AtSGR1/2 genes by BnABI3 is mediated mainly by increased transcription.

The freezing-induced green seed phenotype was rescued in BnABI3 transgenic plants, suggesting that the exogenous expression of BnABI3 might improve freezing tolerance at the seedling stage. To examine this possibility, we compared the development of freezing tolerance after cold acclimation. The BnABI3 transgenic plants acclimated more quickly than the WT plants, while a lag before freezing tolerance started to increase in the abi3 mutants, so the maximum tolerance achieved in the BnABI3 transgenic plants was higher than in the WT (Figure 6D). We checked the expression profiles of SRG1 and SRG2 genes during cold acclimation in WT and BnABI3 plants rosette leaves. We found that SRG1 and SRG2 genes expressions increased to higher level than that in WT control during cold acclimation. We also identified the freezing tolerance without cold acclimation; the results showed that BnABI3 overexpression also increased freezing tolerance without cold acclimation (Supplementary Figure S5). Overall, we showed that SRG1 and SRG2 could play a role in BnABI3 promoted freezing tolerance in transgenic Arabidopsis.

Because the previous study showed that the abi3 mutant exhibits early flowering. Hence, we calculated the 35S::BnABI3 plants flowering time to determine whether BnABI3 regulated the flowering time. The results showed that BnABI3 overexpression delayed flowering time (Figure 8A). There were more rosette leaves at bolting in the 35S::BnABI3 transgenic lines compared with WT (Figure 8B). We also analyzed mRNA level of the flowering time related genes FT, AP1, FLC, and FLM. FT and AP1 showed down regulated expression levels in the 35S::BnABI3 transgenic plants. But FLC and FLM mRNA level increased in 35S::BnABI3 transgenic plants (Figure 8C). The altered expression levels of flowering gene in 35S::BnABI3 plants may contribute to the late flowering phenotype.

In addition to investigating how BnABI3 regulates freezing tolerance, we also investigated the biological function of the ABI3 in the ER stress response. Our results showed that AtABI3 down-regulation confers ER stress tolerance in Arabidopsis. Both gs1 and abi3-6 mutant plants grew as well as Col-0 under our normal growth conditions, but they were more tolerant to ER stress than WT in the medium added with the ER stress inducers tunicamycin (TM) (Figure 9A). Under ER stress conditions, a greater number of G-B (Greenish Big) plants and fewer Y-S (Yellowish Small) plants were observed in the abi3 mutants than in the control (Figure 9B). Taking these findings together, we conclude that mutation of AtABI3 in Arabidopsis promotes ER stress tolerance. Moreover, exogenous expression of the BnABI3 gene under the AtABI3 promoter decreased ER stress tolerance to WT levels (Figure 9A,B). Furthermore, the levels of the ER-resident molecular chaperone genes BIP1, BIP3, and PID1 were up-regulated in gs1 and abi3-6 mutant seeds (Fernandez-Bautista et al., 2017; Iwata et al., 2018) (Figure 9C), and the up-regulated expression of BIP and PDI genes could improve ER stress tolerance. In addition, we investigate the expression levels of ER stress related genes by using BnABI3GR transgenic plants. As shown in Supplementary Figure S6, induced expression of BnABI3 by DEX treatment down regulated the BIP1, BIP3, and PID1 expression levels and result in downregulated stress tolerance. So, we get the conclusion that BnABI3 contributed directly to ER-stress response.

In addition, we want to identify whether ABI3 plays roles in root development. The density of LR in the gs1 and abi3-6 mutants were decreased than that in Col-0, but the density of LR observed in 35S::BnABI3 were increased than that in Col-0 (Figure 10A,B). Because, both mechanical and/or gravitropic stimuli can facilitate the initiation of LR. Previously, reports showed that LR initiation occurs in a highly synchronized manner at the outer surface of a bending root using 90° gravitropic stimulus assays (Peret et al., 2012; Voss et al., 2015). We can classify this developmental progress into many different stages (Figure 10C). In stage I, the LR primordia were observed at 16 h post-gravitropic-induction (pgi), until the LR emerged at 36 h pgi in stage VIII (Figure 10D). Next, WT, mutant and 35S::BnABI3 plants were exposed to a gravitropic stimulus. WT plants accumulated stages I and II LRP at 16 h pgi and stages III to VIII at 36 h pgi. LR initiation and first divisions were not altered in the 35S::BnABI3 plants but altered percentage of stages III–VIII LRP at 36 h pgi (Figure 10D). This result indicated that BnABI3 overexpression affected LR emergence.

The ABA signaling pathway and auxin interact to regulate many different development processes (Brady et al., 2003; Nag et al., 2005; Rock and Sun, 2005). Therefore, we decided to identify the ability of abi3 mutants and 35S::BnABI3 plants root growth in the presence of IAA and the auxin transport inhibitor NPA (N-1-naphthylphthalamic acid) to search for any phenotypes that arise under conditions in which auxin-regulated LR development is perturbed. Under these growth conditions, bs1 and abi3-6 plants required higher concentrations of IAA to initiate LR growth, and 35S::BnABI3 plants increased response to IAA treatment (Figure 10E). Similarly, application of the auxin transport inhibitor NPA inhibited LR initiation in both the WT and abi3 plants; the mutant roots were still less responsive but 35S::BnABI3 plants more response to NPA-induced repression of LR initiation (Figure 10F). Thus, it appears that ABI3 function is required for appropriate auxin responsiveness in the LRs of Arabidopsis plants.