No specific permits were required for the described field studies. The location is not privately owned or protected in any way, and the field studies did not involve endangered or protected species.

Yliangyou 2 (Y58S × Yuanhui 2) hybrid rice (Oryza sativa L.) seeds were a generous gift from Hunan Hybrid Rice Research Center (Changsha, Hunan, China). Water content and germination of seeds were 10.7 ± 0.1% (on a fresh weight basis) and 87%, respectively. The seeds were stored in paper bags at −20°C for up to one year without detrimental effects.

The water content of seeds was gravimetrically determined (at 80°C for 48 h). Four replicates of 25 seeds each were used for determination of water content, and the water content of seeds is expressed on a basis of fresh weight.

To obtain different aging (vigor) levels of seeds, three replicates of 2000 hybrid rice seeds each were stored (aged) in a closed container with 100% relative humidity (RH) and at 40°C for 0, 5, 10, 15, 20, and 25 days, respectively. Water content of seeds increased during artificial aging, and that of seeds artificially aged for 0, 5, 10, 15, 20, and 25 days was 10.7 ± 0.1%, 13.9 ± 0.1%, 14.1 ± 0.0%, 15.1 ± 0.2%, 15.7 ± 0.4%, and 15.9 ± 0.3%, respectively. To decrease the effect of dehydration on the physiological state of the seeds, they were not dried back to the original water content. The seeds were immediately sampled for germination and proteomic analysis after aging.

Three replicates of 50 seeds each were germinated on two layers of filter paper moistened with 4 ml of MilliQ water in closed 90-mm diameter Petri dishes at 25°C in darkness for 168 h. Germination of seeds was checked every 12 h, and the protrusion of a 2 mm radicle was used as the criterion for completion of germination.

The seeds aged for 0, 10, and 25 days, respectively, were manually dehulled and carefully separated into embryo and endosperm, which were immediately frozen in liquid nitrogen. After that, 100 embryos or endosperms were ground to a fine powder in liquid nitrogen with mortar and pestle, and the powder was then kept at −80°C until used.

For extraction of soluble proteins, three replicates of about 0.1 g embryo or 0.75 g endosperm powder each were homogenized in 1.5 ml of precooled extraction buffer contained 50 mM Tris-HCl (pH 7.5), 30% (w/v) sucrose, 10 mM ethylene glycol-bis (β-aminoethylether)- N,N,N′,N′-tetraacetic acid, 1 mM phenylmethylsulfonyl fluoride, 1 mM dithiothreitol (DTT), and 1% (v/v) Triton X-100. After washing the mortar in 0.5 ml extraction buffer, the total of 2 ml homogenate was mixed by vortexing and centrifuged at 4°C at 16,000 g for 10 min, and the supernatant was then centrifuged at 4°C at 32,000 g for 20 min. The resultant supernatant was mixed with two volumes of ice-cold 50 mM Tris-HCl (pH 7.5)-saturated phenol and shaken on ice for 30 min. After centrifugation at 16,000 g for 20 min, the phenol phase was collected and five volumes of precooled methanol saturated with (NH₄)₂SO₄ was added to precipitate the proteins by an overnight incubation at −20°C. The pellets were rinsed four times in ice-cold acetone containing 13 mM DTT, and then lyophilized.

The protein concentration was assayed as described by Bradford (1976) using bovine serum albumin as the standard.

Isoelectrofocusing (IEF) was performed using a Multiphor II horizontal electrophoresis system (Bio-Rad, Hercules, CA, USA) and 17 cm Immobiline Dry Strips with a linear pH gradient of 5–8 (Bio-Rad). Protein sample was loaded onto the strip soaked in re-hydration solution composed of 7 M urea, 2 M thiourea, 4% (w/v) CHAPS, 20 mM DTT, and 0.5% (v/v) immobilized pH gradient buffer (pH 5–8) at 20°C for 12 h. IEF was then performed by applying a voltage of 250 V for 1 h, ramping to 500 V over 1 h, 2000 V for 2 h, and finally 10,000 V until a total of 60 kVh was reached. Prior to the second dimension, the gel strips were equilibrated for 15 min in equilibration buffer containing 50 mM Tris-HCl (pH 8.8), 6 M urea, 30% (v/v) glycerol, 2% (w/v) SDS, 2% (w/v) DTT or 2.5% (w/v) iodoacetamide. After equilibration, the strips were applied to vertical SDS-polyacrylamide gels (5% stacking and 12% resolving), the low-molecular-range markers (Bio-Rad) were loaded at one end of the strip, and then they were sealed with 0.5% (w/v) low-melting agarose in SDS buffer contained 0.01% (w/v) bromophenol blue. After solidification of the agarose, electrophoresis was performed at 15°C in SDS electrophoresis buffer (pH 8.3) composed of 25 mM Tris base, 192 mM glycine and 1% (w/v) SDS, at 25 mA for 30 min and at 40 mA for 4 h. The gels were stained overnight with 0.25% (w/v) Coomassie brilliant blue R-250 (CBB) in 5:1:4 (v/v) methanol: acetic acid: water, and destained with 2:1:7 (v/v) methanol: acetic acid: water solution with several changes, until a colorless background was achieved.

The 2-D gels were scanned at a 300 dpi resolution in a UMAX Power Look 2100XL scanner (Maxium Tech., Taipei, China). Spot detection and gel comparison were made with ImageMaster 2D Platium (version 5.01; GE Healthcare Bio-Science, Little Chalfont, UK). After automated detection and matching, manual editing was carried out to correct the mismatched and unmatched spots.

Three well-resolved gels of each sample were used to create “replicate groups.” The spots, which were well resolved in all three biological replicates, were considered as reproducible (Supplementary Figures S1, S2). The normalized volume of each spot was assumed to represent the abundance of the detected protein. A criterion of ≥2-fold change (2-fold increase/decrease, P < 0.05) was used to define significant differences when comparing spot size among groups.

Protein spots changing in volume were excised from the stained gels. In-gel digestion and tryptic peptide extraction were performed according to the following protocol. Excised spots were washed and destained using a series of washes consisting of 100 μl of water, 100 μl of 50% (v/v) acetonitrile (Fisher Scientific; Fair Lawn, NJ, USA) and 50 μl of 100% acetonitrile. The proteins in the gel pieces were reduced with 10 mM DTT in 100 mM NH₄HCO₃ at 56°C for 45 min, and then incubated with 55 mM iodoacetamide in 100 mM NH₄HCO₃ at 20°C in darkness for 30 min. After removing iodoacetamide and a series of washes described above, the gel pieces were rehydrated in 50 mM NH₄HCO₃ with 10 ng trypsin (sequencing grade modified, Promega, Madison, WI, USA) on ice for 45 min and then incubated overnight at 37°C. After digestion, the supernatant from each sample was recovered and the remaining peptides were then extracted using 5 μl of 5% (v/v) trifluoroactic acid (TFA) followed by 50 μl of 50% (v/v) acetonitrile with 2.5% (v/v) TFA. Each sample was sonicated for 5 min before collecting the supernatant. All supernatants were combined from each spot. After desalting using a column of POROS R2 resin, the samples were spotted on a MALDI target plate, and immediately spotted on top with 0.5 μl of saturated matrix (containing 10 mg/ml α-cyano-4-hydroxycinnamic acid (Sigma), 50% (v/v) acetonitrile, 0.1% (v/v) TFA and 1 mM ammonium phosphate), and dried completely. Samples were then subjected to MALDI-TOF/TOF MS analysis (UltrafleXtreme, Bruker Daltonics, Bremen, Germany).

The peptide mass fingerprints obtained were searched against Oryza sativa in the SwissProt and/or NCBI database using MASCOT software (Matrix Science, London, UK). The following search parameters were applied: SwissProt and/or NCBI were used as the protein sequence database; a mass tolerance of 70 ppm in MS mode and 0.5 Da for MS/MS and one incomplete cleavage were allowed; acetylation of the N-terminus, alkylation of cysteine by carbamidomethylation, and oxidation of methionine were considered as possible modifications. The proteins had to meet the following criteria: (1) the probability-based MOWSE score of identified proteins was greater than 48 (P < 0.05) in SwissProt database and it was greater than 64 (P < 0.05) in NCBI database; (2) the number of matched peptides was at least two.

All data were analyzed with SPSS for Windows 19.0 (SPSS Inc., 2010). Changes in seed germination and accumulation ratios and associated P-values for differentially changed proteins in embryos and endosperms during artificial aging of Yliangyou 2 hybrid rice seeds were analyzed using a one-way ANOVA followed by Student-Newman-Keuls multiple comparisons test (S-N-K, P = 0.05). All values are expressed as means ± SE (seed germination) or means ± SD (the volume of differentially changed spots).

Germination of Yliangyou 2 hybrid rice seeds, an important parameter to assess seed viability, significantly decreased during aging at 100% RH and 40°C. After 12 days, 50% of the seeds had lost germinability, while none of the seeds aged for 25 days were able to germinate (P ≤ 0.001, Figure 1A). We observed that with increasing aging time, the seed germination rate, another important parameter to assess seed viability, also decreased. For example, the time taken to reach 50% germination was about 42, 56 and 80 h, respectively, for seeds aged for 0, 5, and 10 days, and 50% germination was never reached for seeds aged for 15 and 20 days (Figure 1B).

To identify the proteins involved in hybrid rice seed aging, the total protein content of embryos and endosperms from the seeds aged for 0 (A), 10 (B), and 25 (C) days was extracted and analyzed by 2-DE (Figure 2). By comparing the protein profiles of different seed samples, 1109 ± 103 and 1093 ± 93 protein spots were detected in embryo and endosperm, respectively, during aging of Yliangyou 2 hybrid rice seeds (Supplementary Figures S1, S2). Among these protein spots, a total of 91 and 100 protein spots showed a significant change in abundance (≥2-fold increase/decrease, P < 0.05) in embryos and endosperms, respectively (Supplementary Tables S1-S4, Figure 3). All of protein spots changing in volume were excised and analyzed by MALDI-TOF/TOF MS, and were identified by first searching in the SwissProt database and, if that showed no matches, then in the NCBI database.

Of 91 protein spots changing in volume in embryos, 71 spots gave one significant protein match, and four spots gave at least two protein matches (Table 1, Supplementary Tables S1, S3, S5, S7, Figure 3A); out of 100 protein spots changing in volume in endosperms, 79 spots gave one significant protein match, and 13 spots gave at least two protein matches (Table 2, Supplementary Tables S2, S4, S6, S8, Figure 3B); when searching in the SwissProt and/or NCBI database. Since we do not know which of the two or more than two proteins showed a significant change in abundance in four protein spots in embryos and 13 protein spots in endosperms among different seed samples (Supplementary Tables S7, S8), these 17 protein spots were not considered in the following analysis.

The identified 71 single protein spots in embryos were matched to 51 unique genes, and could be classified into seven functional groups and 18 subfunctional groups based upon Bevan et al. (1998) and Schiltz et al. (2004) (Table 1, Figure 4A). Most of the identified proteins were associated with energy (30%), with cell defense and rescue (28%), with storage protein (18%), with protein synthesis and destination (10%), and with metabolism (9%); these proteins accounted for 95% of all identified proteins in the embryos (Figure 4A). Furthermore, the identified 79 single protein spots in endosperms were matched to 60 unique genes, and could be classified into seven functional groups and 21 subfunctional groups (Table 2, Figure 4B). Of these proteins, 37% protein spots were related to metabolism, 27% to energy, 11% to protein synthesis and destination, 9% to cell defense and rescue, 6% to storage protein, and 6% to cell growth and division; these proteins accounted for 96% of all identified proteins in the endosperms (Table 2, Figure 4B).

By comparing the proteome profiles of embryos and endosperms from seeds aged for 0, 10, and 25 days, respectively, the proteins changing in abundance can be classified into those with increased/decreased abundance and those with a complex abundance pattern (Tables 1–3, Supplementary Tables S3, S4, Figure 3). The increased proteins are those whose abundance, compared to 0 day, increased by ≥2-fold at 10 days of aging and stayed high or increased further at 25 days of aging, and those whose abundance showed a less than 2-fold change at 10 days of aging and then a ≥2-fold increase. The decreased proteins are those whose abundance, compared to 0 day, decreased by ≥2-fold at 10 days of aging and stayed low or decreased further at 25 days of aging, and those whose abundance showed a less than 2-fold change at 10 days of aging and then a ≥2-fold decrease. The proteins whose abundance first increased by ≥2-fold and then decreased significantly (including less than 2-fold change) or vice versa during seed aging are considered to show a complex pattern (Tables 1–3, Supplementary Tables S3, S4, Figure 3).

During seed aging, 61 embryo protein spots showed an increase/decrease in volume, and 10 spots showed a complex pattern; while 77 endosperm protein spots showed a decrease/increase in volume, and two spots showed a complex pattern (Table 3, Supplementary Tables S3, S4, Figure 3). Of the 61 embryo protein spots, 58 increased and three decreased in abundance. Out of the increased proteins, 19 were involved in energy, 15 in cell defense and rescue, 11 in storage protein, five in metabolism, five in protein synthesis and destination, two in transcription, and one in cell growth and division. The three embryo proteins decreasing in abundance were associated with energy, transcription and cell defense and rescue (Table 3, Supplementary Table S3, Figure 3A).

Out of 77 changing protein spots in endosperms, 76 decreased in volume while only one increased. Among the proteins decreasing in abundance, 29 were involved in metabolism, 21 in energy, eight in protein synthesis and destination, seven in cell defense and rescue, five in storage protein, three in cell growth and division, and three in unknown protein; for the increased protein, only one protein was implicated in cell growth and division (Table 3, Supplementary Table S4, Figure 3B).

When hybrid rice seeds were exposed to 100% RH at 40°C, germination percentage and germination rate of seeds significantly decreased (Figure 1), indicating that seed viability decreased during artificial aging. These results are similar to those of Rajjou et al. (2008), Xin et al. (2011), Zhang et al. (2015) and Gao et al. (2016), who found that seed viability gradually decreased with increasing aging time. In this way, we obtained seed samples with different viability (aging level).

To understand the action of embryo and endosperm and their interaction in seed aging, the differentially changing proteins in embryo and endosperm were compared in hybrid rice seeds with different aging levels. Using a ≥2.0-fold change (P < 0.05) in abundance as the significance level, we found that during aging of hybrid rice seeds, 95% of the changing proteins increased in embryos, while 99% decreased in endosperms (Table 3, Supplementary Tables S3, S4, Figure 3). Most of the differentially changing proteins were embryo- or endosperm-specific, 14 proteins (spots 10, 11/18′, 19′; 22/31′; 13/23′; 47, 49/75′; 14/26′; 35/43′; 16, 17, 18, 20/28′, 29′, 30′; 48/73′, 74′; 6, 7/14′; 84/93′; 29, 30, 61/79′; 71, 75/96′; 12/21′; 74/86′) were commonly present in both embryo and endosperm (Tables 1-3, Supplementary Tables S3, S4, Figure 3), although their abundance changes were opposite. These results indicate that the embryo and endosperm proteomes were differentially affected by aging.

It is possible that seed aging is an active process, in which the embryo tries to adjust to the aging conditions by synthesizing a range of proteins, while the endosperm provides amino acids and related products for protein synthesis in the embryo by degrading proteins. The endosperm is considered to be a “dead” tissue in the sense that, except for the aleurone layer, there are no cells that can grow and divide (Bewley et al., 2013). Any change in the starchy endosperm dependent on changes in gene expression therefore requires some form of communication with the aleurone layer and/or the embryo. The induction of α-amylase biosynthesis in the aleurone layer by gibberellic acid produced in the embryo during the early stages of germination is an example of this (Bewley et al., 2013). Han et al. (2014) reported a much broader regulation of endosperm metabolism by the embryo in germinating rice seeds. Our results indicate that a similar regulation of endosperm metabolism by the embryo could take place during aging of rice seeds.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.