High seed quality is a key trait to achieve successful crop establishment required for optimum yield and sustainable production. Seed storage conditions greatly impact two key seed quality traits; seed viability (ability to germinate and produce normal seedlings) and vigour (germination performance). Accumulated oxidative damage accompanies the loss of seed vigour and viability during ageing, indicating that redox control is key to longevity. Here, we studied the effects of controlled deterioration at 40°C and 75% relative humidity (RH) (‘ageing’) under two different O2 concentrations (21 and 78% O2) in Brassica oleracea. Two B. oleracea genotypes with allelic differences at two QTLs that result in differences in abscisic acid (ABA) signalling and seed vigour were compared. Ageing led to a similar loss in germination speed in both genotypes that was lost faster under elevated O2. In both genotypes, an equal oxidative shift in the glutathione redox state and a minor loss of α-tocopherol progressively occurred before seed viability was lost. In contrast, ABA levels were not affected by ageing. In conclusion, both ABA signalling and seed ageing impact seed vigour but not necessarily through the same biochemical mechanisms.
Seeds age during storage and the resulting loss of seed vigour and viability can impact crop yields and sustainable agricultural production. Seed deterioration rates are accelerated by elevated storage temperatures and seed water contents (WC), which can also be influenced by the O2 content of the storage environment [1–5]. Vigorous germination, resulting in rapid and uniform seedling establishment, is an important agronomic trait . Plant hormones are determining factors for seed germination and vigour, with prominent roles for abscisic acid (ABA) and gibberellins (GA, gibberellic acid) that can delay and advance germination, respectively . In seeds, ABA has a well-characterised role in maintaining dormancy, whereby germination is restricted in otherwise favourable conditions [7,8]. In addition, ABA signalling is important in coordinating correct embryo development for producing high-quality seed. Arabidopsis mutants that are ABA insensitive or ABA-deficient display reduced seed longevity . An inverse relationship between the level of seed dormancy and longevity in various inbred Arabidopsis lines was revealed , but this link is not consistent in Arabidopsis, nor was it found in sunflower seeds [11,12].
Oxidative processes affect germination, whereby an ‘oxidative window of germination’ has been described . On the one hand, applications of reactive oxygen/nitrogen species (ROS/RNS) are often used to break seed dormancy and promote germination, via cross-talk with ABA and GA signalling [14,15]. On the other hand, aged seeds with low germination display symptoms of excess ROS production, such as carbonylated proteins and depleted antioxidant pools [16,17]. A loss of the antioxidant and redox buffer glutathione (GSH), alongside an accumulation of disulfides (i.e. GSSG [glutathione disulfide]), occurs during seed ageing, resulting in an oxidative shift in the GSH half-cell reduction potential (EGSSG/2GSH) [11,17–19]. Glutathione reductase (GR) converts GSSG to GSH in the presence of NADPH, but only when seeds are sufficiently hydrated. GSH is the most abundant low-molecular-weight (LMW) thiol and an important cellular redox buffer , particularly in desiccated seeds that are devoid of ascorbate [21,22]. Seed viability in the Brassicaceae family has also been associated with antioxidant protection of the lipid phase by tocochromanols [12,23], the collective name for tocopherols and tocotrienols of the vitamin E family. However, a decrease in tocochromanols does not always occur in aged seed [11,24–26].
Several important crops, including cabbage, kale, kohlrabi, cauliflower, broccoli and Brussel sprouts, belong to the same species, Brassica oleracea L., the nearest crop relative to the model plant Arabidopsis thaliana L. According to the Food and Agriculture Organization (FAO), 3.8 million hectares of cabbage, cauliflower and broccoli were harvested in 2016, corresponding to a global production of 96.5 million tonnes (www.fao.org/faostat/). Previously, a B. oleracea mapping population was investigated, containing lines that had major differences in their reproductive biology . An investigation into the variation of germination vigour in these lines identified quantitative trait loci (QTL) underlying this trait . Further investigation showed that three genes (BoLCVIG1, BoLCVIG2 and RABA1) were responsible for differences in seed vigour within B. oleracea. Two lines, namely A12DHd and AGSL101, were identified that differed in seed vigour due to allelic differences in these genes [29,30]. This allelic variation altered both seed ABA content and sensitivity to it, affecting the speed of germination. Germination vigour includes all aspects of seed performance. In addition to the initial vigour that is genetically determined and hormone-related, vigour is also affected by seed deterioration that reduces performance. The latter results in a decline in the same germination measures used to indicate initial seed vigour. Therefore, these two aspects can be confounded in research and in commercial estimates of vigour. Here, we used A12DHd and AGSL101 to investigate if genetically determined, ABA-related, differences in initial seed vigour influence mechanisms of seed ageing and ageing-induced losses of seed vigour. Seeds were aged with controlled deterioration (CD) at 40°C and 75% RH, either under ambient (21%) or elevated (78%) O2 to further assess the influence of an oxidative environment. Changes in germination rate, levels of LMW thiols and disulfides, plant hormones, tocochromanols and fatty acids (FAs), and GR activity were measured before viability loss (0, 4 and 8 days of CD), when total germination (TG) started to be lost (19-day CD), and when ∼50% of TG was lost (41-day CD).
Seed material and germination
Seeds of the slow-germinating B. oleracea double haploid Chinese kale parental line (var. alboglabra A12DHd; hereafter, A12) and a chromosome substitution line (AGSL101; hereafter, SL101) in the A12 background with introgressions from a fast-germinating doubled haploid Calabrese line (var. italica GDDH33) at two seed performance QTLs were used. GD33 alleles enhance seed performance resulting in more rapid germination and seedling emergence in the field . For seed production, A12 and SL101 plants were grown in a glasshouse and pollinated as previously described . Desiccated whole seeds, with a WC of 6.1 ± 1.1% on a fresh weight basis, were stored at −20°C before ageing by CD. Seeds were germinated on moistened germination paper (Whatman grade 3644, GE Healthcare, U.K.) in the dark at 16°C. A seed was considered germinated when radicle emergence was greater than 2 mm.
CD and seed preparation for biochemical analyses
Seeds were pre-equilibrated at 70% RH over non-saturated LiCl for 2 days at room temperature to establish a seed WC of 7.7 ± 0.2%, before CD at 40°C and 75% RH over saturated NaCl solution to maintain the same seed WC. Four replicate seed lots were aged in separate sealed boxes containing ambient O2 levels (18–21%) or elevated O2 levels (75–78%). The O2 concentrations were monitored in 1 ml of H2O contained in the box and readjusted weekly using an Oxygen Dipping Probe connected to a Fibox 3 interface (PreSens, Germany). After various intervals, seeds were removed for germination testing or immediately frozen in liquid nitrogen and stored at −80°C for biochemical analyses (n = 20 seeds per replicate). Thereafter, seeds were freeze-dried for 3 days. Freeze-dried seeds (20 seeds = c. 100 mg) were ground in a 2 ml Eppendorf tube with a 8 mm metal bead for 10 min at 25 Hz (TissueLyser II, Qiagen).
LC–MS/MS analysis of ABA and salicylic acid
Analyses of ABA and salicylic acid (SA) were carried out according to ref. . In brief, 20 mg of lyophilised seed powder was extracted in 1.5 ml of ice-cold acetone/water/acetic acid (80:20:1, v:v:v) after the addition of 25 µl stable isotopically labelled internal standard solution (1 µM ABA-d6, 5 µM SA-d4) by shaking (TissueLyser II, Qiagen, Düsseldorf, Germany) at 30 Hz for 5 min using one 5 mm glass bead per Eppendorf tube, followed by centrifugation at 10 000×g, 4°C for 12 min. Supernatants were evaporated to dryness using a SpeedVac SPD111 vacuum concentrator (Thermo Fisher Scientific Inc., Waltham, MA, U.S.A.), followed by resuspension in 150 µl of ACN/water (50:50, v:v), supported by 5 min ultrasonication in an ice-cooled water bath. The extracts were filtered through 0.2 µm PTFE filters before injection into the UHPLC–MS/MS system. ABA and SA were identified and quantified by LC–MS, using an ekspert ultraLC 100 UHPLC system coupled to a QTRAP 4500 mass spectrometer (AB SCIEX, Framingham, MA, U.S.A.).
HPLC analysis of LMW thiols and disulfides and tocochromanols
LMW thiols and disulfides were determined by HPLC as described in . Briefly, LMW thiol/disulfide samples were extracted from 10 mg of freeze-dried and ground seeds in 0.1 M HCl containing 10 mg polyvinylpolypyrrolidone and 0.5% (v:v) Triton X-100. Samples were sonicated for 15 min and then centrifuged for 20 min at 15 000×g at 4°C. The supernatant was collected, avoiding the lipid and pellet phases and centrifuged again for 20 min at 15 000×g at 4°C. Thiols were labelled by monobromobimane (mBBr) for fluorescence detection (excitation: 380 nm; emission: 480 nm) and disulfides were determined after blockage of thiols with N-ethylmaleimide and reduction in disulfides with DTT. The half-cell reduction potential of LMW thiols was calculated from the concentrations of thiols and disulfides considering the seed WC (g water per g seed dry weight [DW]) to calculate molar concentration and taking into account deviations from standard conditions in terms of pH (=7.3) and temperature (=298 K) using the Nernst equation according to  and .
Tocochromanols were analysed by HPLC following the procedure described in  with slight alterations. Briefly, 20 mg of freeze-dried and ground seeds was extracted three times in 1 ml ice-cold heptane by vortexing for 20 s and centrifugation at 13 000×g for 20 min at 4°C. The combined supernatants were re-centrifuged at 13 000×g for 20 min at 4°C prior to separation on a reversed-phase column (LC-Diol, 250 × 4.6 mm i.d., 5 µm particle size, Supelco Analytical, Supelcosil™) and detection (fluorescence excitation: 295 nm, emission: 325 nm). To quantify high and low abundant tocochromanols, 10 and 70 µl of extract were injected, respectively, that were identified and quantified against individual tocopherol and tocotrienol standards.
Colorimetric analysis of GR activity
30 mg of freeze-dried and ground seed material was suspended in 1 ml of ice-cold hexane to remove interfering lipids and centrifuged at 6000×g for 15 min. After discarding the hexane, the pellet was rinsed in fresh hexane and air-dried before the seed powder was dissolved in 700 µl Sorensen's phosphate buffer (pH = 7.8), containing a protease inhibitor. The samples were centrifuged at 20 000×g for 15 min at 4°C, the supernatant was collected and re-centrifuged for 15 min at 20 000×g. Proteins were precipitated over 1 h at −20°C with 1.4 ml 100% acetone added directly to the extract. Proteins were collected by centrifugation for 10 min at 6000×g and 4°C and the acetone was discarded. The step was repeated with 80% acetone. After air-drying the protein pellet, it was suspended in 800 µl of Sorensen's phosphate buffer (pH 7.8) and 30 µl protease inhibitor cocktail used immediately. Enzyme kinetics was measured using a PerkinElmer spectrophotometer (Lambda 800 UV/VIS Spectrometer) with a cuvette holder maintained at 25°C. GR was measured by the decrease in absorbance at 340 nm (ε = 6.22 mM−1 cm−1) with 0.083 mM GSSG, 0.083 mM NADPH and 200 µl enzyme extract in a total reaction volume of 600 µl in 0.05 M Sorenson's buffer at pH 7.8.
Gas chromatography–mass spectrometry analysis of FAs
FAs were derivatised to fatty acid methyl esters (FAMEs) as previously described . Briefly, 10 mg of freeze-dried seed powder was treated with 2.64 ml of methanol:toluene:sulfuric acid 10:3:0.25 v:v:v containing 0.01% (w:v) butylated hydroxytoluene. 200 µg of heptadecanoic acid (C17:0, dissolved in hexane) was added simultaneously as the internal standard. Samples were incubated at 80°C for 90 min with constant agitation before adding 1 ml of hexane and 3 ml of 0.9% NaCl (w:v). Samples were vigorously mixed before centrifugation for 10 min at 3000×g. The supernatant was transferred to autosampler vials and kept at −20°C if not analysed immediately. FAMEs were separated using a Trace 1300 gas chromatograph (GC) (Thermo Fisher Scientific Inc., U.S.A.) on a 30 m FAMEWAX column (Restek #12497, Bellefonte, U.S.A.) and detected using a TSQ 8000 triple quadrupole detector (Thermo Fisher Scientific Inc., Waltham, MA, U.S.A.) operated in full scan mode (50–550 m/z). 1 µl of sample was injected in split mode (split ratio 1/100) in a split/splitless (SSL) inlet heated at 230°C and containing a Mini-Lam 4 mm split liner (#453A2009, Thermo Fisher Scientific Inc., Waltham, MA, U.S.A.). The temperature gradient was first set to 120°C for 1 min, then increased by 16°C per minute up to 190°C, then by 5°C per minute up to 220°C and finally increased by 2°C per minute up to 235°C, which was held for 7 min with a carrier gas flow of 1.2 ml min−1 of helium. The ion source temperature was set to 250°C and the transfer line to 240°C. A commercial FAME mix (ref. 18919, Sigma–Aldrich, St. Louis, MO, U.S.A.) was used to confirm the identities of the FAs. External standards of palmitic, stearic, oleic, linoleic and linolenic acid were used in conjunction with the internal standard to estimate the total amount of each FAs. Data analysis was performed using the Xcalibur software (Thermo Fisher Scientific Inc., Waltham, MA, U.S.A.).
Statistical analysis of data was carried out with the SPSS software package (v. 23) via three-way analysis of variance (ANOVA), except FA data that were tested with a Student's t-test. Univariate ANOVA-derived P-values for differences in all data are presented in each figure between genotypes, over time of ageing or in response to the level of O2. For example, when a significant ageing effect is stated (P < 0.05), all data for both genotypes aged under both O2 levels were considered. For assessing significant changes in viability, the number of seeds that had germinated, and not % TG, was considered to avoid the need for data transformation.
Non-aged A12 seeds took 3.4 ± 0.1 days to reach 25% germination (T25%), slower than the 2.5 ± 0.1 days it took SL101 seeds (Figure 1A). Values of T25% increased over the first 19 days of CD during which a significant difference in vigour between the genotypes was maintained (Figure 1A). The ABA levels in seeds of A12 and SL101 were c. 0.5 and 2.5 nmol g−1 DW, respectively, and were not significantly altered during 41 days of CD (Figure 1B). SA significantly accumulated during CD, but no significant effect of oxygen concentration and no difference between genotypes was found (Supplementary Figure S1). The TG remained above 90% until 19 days of CD (Figure 1C), but additional O2 during CD (CD + O2) over this period led to a more rapid ageing-induced loss in vigour, as shown by the increased T25% values (Figure 1A). However, CD + O2 had no consistent negative effect on TG, which dramatically declined between 19- and 41-day CD (Figure 1C), at which point the T25% values were >20 days and the vigour of the genotypes was indistinguishable (Figure 1A). Overall, seeds of genotype SL101 had lower ABA contents and faster germination rates than those of genotype A12. The difference in vigour between the genotypes was maintained during ageing before viability was lost and elevated O2 during ageing accelerated vigour loss.
Ageing-induced changes on ABA levels, germination speed and losses of TG.
Non-aged seeds of both genotypes contained ≈1.75 µmol g−1 DW total GSH, while possessing a GSH:GSSG ratio of ≈2:1 (Figure 2A) and EGSSG/2GSH values of approximately −210 mV (Figure 2B), which is typical for highly viable seeds . A significant decrease in GSH levels occurred in response to CD, while GSSG levels increased (Figure 2A), contributing to an oxidative shift in EGSSG/2GSH, which after 19 days of CD rose to approximately −155 mV. However, no clear effect of CD + O2 on GSH nor GSSG levels was observed (Figure 2B). By 41-day CD, GSH levels were ≈0.2 µmol g−1 DW and the GSH:GSSG ratio was ≈1:5 (Figure 2A), showing that the majority of GSH was converted to GSSG while ≈30% of total GSH was consumed. GR activities in non-aged seeds were ≤14 nkat g−1 DW and were halved after 41 days of CD (Figure 2C), although overall SL101 had significantly higher GR activity than A12 (Figure 2C). Concomitantly with a CD-induced accumulation of GSSG, levels of other LMW disulfides also increased, including cystine and bis-γ-glutamylcysteine, which after 41 days of CD had increased two-fold and four-fold, respectively (Supplementary Figures S2 and S3). In summary, ageing was associated with an oxidative shift in the LMW thiol-based cellular redox state, whereby thiols (i.e. GSH) converted to disulfides (i.e. GSSG), but these changes were not influenced by genotype, nor by elevated O2 during ageing.
Ageing-induced changes in glutathione metabolism.
Four different types of tocochromanols were identified in whole seeds; γ-tocopherol (70%), α-tocopherol (19% A12; 17% SL101), β-tocotrienol (7% A12; 9% SL101) and γ-tocotrienol (4%), which together amounted to 270 ± 10 µmol g DW−1 of total tocochromanols in untreated seeds (Figure 3). There was a significant effect of CD on α-tocopherol, which decreased over time, while γ-tocopherol and γ-tocotrienol levels tended to increase. A significant difference in α-tocopherol, β-and γ-tocotrienol between genotypes was found when considering all time intervals, while CD + O2 significantly lowered β-tocotrienol levels relative to CD under ambient O2. Overall, in response to CD total content of tocochromanols remained stable, but the overall composition changed due to declining levels of α-tocopherol and β-tocotrienol, which was compensated by increasing levels of γ-tocopherol and γ-tocotrienol (Figure 3).
Ageing-induced changes in tocochromanols.
In total, 18 FAs were detected using GC–MS. The abundance of each FA was approximated from total ion current (TIC) chromatograms (Figure 4A), whereby the 10 most abundant FAs, representing >95% chromatogram area of the total FAs, were palmitic acid (16:0), stearic acid (18:0), oleic acid (18:1), linoleic acid (18:2), linolenic acid (18:3), eicosanoic acid (20:0), gadoleic acid (20:1), eicosadienoic acid (20:2), docosanoic acid (22:0) and erucic acid (22:1). Comparing the relative FA levels between genotypes, 16:1, an isomer of 18:1, 18:2, 22:0, an isomer of 22:1 and docosadienoic acid (22:2) were significantly higher in non-aged SL101 than A12 (Figure 4B), indicating a minor effect of ABA signalling during seed development on lipid profiles, but there was no effect of ageing on relative FA contents in either genotype (Supplementary Table S1).
Relative FA contents in the non-aged seed.
Following physiological maturity, seeds begin to deteriorate, whereby vigour declines prior to viability [1,6]. The extent to which the underlying biochemical pathways leading to losses of vigour and losses of viability are shared is unclear. Here, we compared the effects of ageing by CD on seeds of two B. oleracea lines that have allelic differences in key vigour-determining genes , but otherwise, share a common genetic background. These genes determine ABA content at physiological maturity and sensitivity to ABA and therefore, initial seed vigour [29,30]. Seed vigour is a ‘catchall’ term, comprising all aspects of seed performance. Initial seed vigour is genetically determined, as well as being affected by environmental factors during seed development, but is lost due to seed deterioration . These aspects are confounded in many vigour studies and in commercial estimates of vigour. The A12 (low initial vigour) and SL101 (high initial vigour) lines, used in this work, allowed us to study how this genetically determined vigour difference may affect seed deterioration and vigour loss induced by ageing.
Seed deterioration due to seed ageing has been linked with oxidative modifications, including the overall cellular redox state [17,35,36]. Therefore, it could be expected that seeds age faster under high O2 concentrations. Studies using elevated partial pressures of O2 during CD supported this assumption . Within the first 19 days of ageing, prior to the drop in viability, germination speed was slowed by elevated O2 (Figure 1A). Interestingly, the ABA-associated genetic differences that determine the initial speed of germination had no influence on the O2-associated loss of vigour. Elevated O2 during CD also resulted in a loss of vigour in non-dormant Helianthus annuus and Hordeum vulgare seeds [11,25]. However, in this study, as also shown previously for H. annuus  or H. vulgare seeds , elevated O2 during CD had limited impact on viability loss. Other studies support the idea that seed ageing can be influenced by O2 [4,37–39], but viability loss also includes O2-independent factors. One explanation for the inconsistent effects of O2 on seed viability loss is the variability of molecular mobility. Seed WC appears to be decisive for seed responsivity to O2 during ageing. At higher WC, O2 may promote seed longevity by allowing O2-mediated metabolism, whereas decreasing O2 levels increases seed survival at slower ageing rates [3,5,37,40]. The minor effects of O2 levels on viability loss observed here (Figure 1C) indicate that the potentially deteriorative effects of O2 were outweighed by its potential benefit for O2-dependent metabolism.
During ageing, and before viability loss, levels of GSSG rose while GSH contents and GR activity decreased, which equally occurred for both genotypes (Figure 2). It is unlikely that GR is able to efficiently reduce GSSG during CD due to restricted molecular mobility in seed equilibrated to 75% RH. However, decreased GR activity in aged seeds would affect the thiol-based cellular redox state of seeds during imbibition when GSSG is converted back to GSH . Changes in GSH redox state can be expressed through EGSSG/2GSH, a Nernst equation-derived calculation that considers molar concentrations of both GSH and GSSG. The changes in GSH and GSSG amounts during seed ageing led to an oxidative shift of the EGSSG/2GSH by 60 mV to approximately −160 mV during the first 19 days of CD (Figure 2B). A zone between −180 and −160 mV has been associated with a 50% viability loss of a seed population  and EGSSG/2GSH values did not change between 19 and 41 days of CD when viability dropped to ≈50%. In both genotypes, a shift of EGSSG/2GSH to less negative values appears to be associated with the loss of seed vigour (Figure 1A). However, the weak significant effect of O2 on EGSSG/2GSH values during CD (P = 0.28) is insufficient to be able to directly associate the enhanced O2-associated decrease in germination vigour with an oxidative shift of EGSSG/2GSH.
The differences in FA and tocochromanol composition between the genotypes (Figures 3 and 4) indicate that ABA signalling can influence lipid-related pathways. In Medicago truncatula, an interconnection between tocochromanol synthesis and ABA signalling has been demonstrated [42,43]. Tocochromanol-deficient mutants of A. thaliana were not able to retain viable embryos  and incurred an elevated level of lipid peroxidation during seed storage, germination, and early seedling development . In A. thaliana and B. oleracea, γ-tocopherol is the dominant tocochromanol, but in this study levels of γ-tocopherol did not decline in response to CD (Figure 3A). In contrast, an increase in tocochromanols in both genotypes occurred within the first 4 days of CD, indicating either tocochromanol synthesis or conversion from existing precursors. A lack in change of FA profiles of aged seed (Supplementary Table S1) indicates that polyunsaturated fatty acids were well protected from peroxidation during CD. The second most abundant tocochromanol, α-tocopherol, was significantly affected by CD in both genotypes, decreasing after 8 days (Figure 3B). This different response of γ- and α-tocopherol is consistent with the faster degradation of α-tocopherol than that of γ-tocopherol observed in seeds of Brassica napus , Pinus sylvestris  and Suaeda maritima , suggesting that either α-tocopherol is more efficient at radical scavenging than γ-tocopherol or that it is located in lipid domains more affected during seed ageing.
Summary and conclusion
As summarised in Figure 5, the differences in ABA levels and signalling between the investigated genotypes had a significant impact on germination speed, which were maintained during CD before viability was lost. Other more subtle genotype-associated differences in tocochromanols, FA profiles and GR activity were found. However, these had no apparent impact on rates of seed ageing and associated biochemistry; a loss of GSH, α-tocopherol and GR activity, an oxidative shift in EGSSG/2GSH and an increase in SA all occurred to a similar extent in both B. oleracea genotypes. Therefore, low seed vigour due to ABA signalling is based on different biochemical mechanisms from those resulting from deterioration.
Summary of the seed properties associated with ABA signalling and ageing in B. oleracea.
Brassica oleracea line A12DHd
analysis of variance
- CD + O2
seeds aged with CD under 75% O2
glutathione half-cell reduction potential
fatty acid methyl esters
Food and Agriculture Organization
gas chromatography–mass spectrometry
quantitative trait loci
reactive oxygen/nitrogen species
Brassica oleracea line AGSL101
time taken to reach 25% germination
T.R. and I.K. conceived the study. All authors contributed to the experimental design. C.S., T.R., W.S. and E.A. conducted the experiments and T.R. analysed the data. All authors wrote the manuscript.
We acknowledge funding from the Seventh Framework Programme of the European Union ‘EcoSeed’ (Impacts of Environmental Conditions on Seed Quality) under grant number .
We are grateful to Siegfried Aigner for the technical assistance (Department of Botany, University of Innsbruck, Austria), and to Katherine Dent for the production of seeds (School of Life Sciences, University of Warwick, U.K.).
The Authors declare that there are no competing interests associated with the manuscript.