The study included two sets of experimental seed samples: I) 20 different rice seed lots of different accessions were used in an experiment to study their aging under EPPO conditions, and II) seven other seed lots to compare their aging under different storage conditions ( Supplementary Table 1 ). For the first experiment, the 20 seed lots were provided by the T.T. Chang Genetic Resources Center, International Rice Research Institute (IRRI), Philippines. These seeds were produced during the 2013 dry season, dried in the gene bank drying room (15% RH and 15°C) over 14 days, and then stored at 2-4°C in sealed laminated aluminum foil packets (Whitehouse et al., 2015). For the second experiment, comparing EPPO aging with natural aging and CD conditions, seven seed samples were taken from commercial seed lots produced in India during the wet season of 2014 and 2015. The seeds of varieties Thanu, IR64, and Raksha were obtained from the Zonal Agricultural Research Station (ZARS), Mandya, India, while those of varieties PB 2511 and PB 1509 were obtained from the Seed Production Unit, ICAR-Indian Agricultural Research Institute, New Delhi, India. Seed lots produced in 2014 were stored for seven months in a seed warehouse in India without any control for temperature and RH. In contrast to the usual procedure, the seed samples used in this study neither received a hot water treatment nor chemical treatments for eliminating potential nematodes or fungal pathogens because such treatments may interfere with the seed aging assays. Upon arrival in the laboratory at Wageningen University & Research (WUR), The Netherlands, the water activity of seed samples was determined (see below). Seeds were equilibrated for 14 days in a dedicated storage cabinet with circulating air maintained at 30% RH and 20°C. After this equilibration, seed samples were hermetically sealed inside laminated aluminum foil packets and stored at 20°C until use in the various experiments. A small quantity of seeds from each seed lot were also stored at both -28°C and -80°C for later use.

Water activity (a_w) of at least 100 seeds per sample was measured at approx. 20°C using an HC2-AW probe connected to a HygroLab 3 display unit (Rotronic Measurement Solutions, Switzerland). The equilibrium relative humidity (ERH) is the RH of the air surrounding the seed that is in equilibrium with its environment. When the equilibrium is achieved, the ERH (%) was subsequently calculated by multiplying the a_w by 100 (Kong and Singh, 2016).

Oxygen levels were measured using an optical oxygen sensor spot (Pst3 type, PreSens Precision Sensing GmbH, Germany, https://www.presens.de), placed against the inner wall of transparent glass storage jars, in conjunction with a fiber optic oxygen meter (PreSens Precision Sensing GmbH). Oxygen readings were recorded at room temperature (RT; approx. 20°C). Before measurement, the oxygen meter was calibrated using an oxygen sensor spot placed inside a jar filled with a sodium azide solution (0% oxygen) and with fresh air (21% oxygen). The PreSens oxygen measurement showed slight fluctuations in the reading (< 1%), hence the lowest of three sequential readings was recorded as the actual oxygen level.

Seed germination assays were performed as described in the International Rules for Seed Testing (ISTA, 2018) with slight modifications. Two replications of 40-45 seeds per treatment were used in a randomized complete block design. Dry seeds were sown on two layers of dry filter paper (142 mm × 203 mm (L × W) blue blotter paper; All Paper, Zevenaar, the Netherlands) placed in plastic trays (150 mm × 210 mm (L × W); DBP Plastics, Antwerpen, Belgium). Germination was initiated by dispensing 50 mL of demineralized water to each tray. Watered trays were stacked along with one empty watered tray, each at the bottom and on the top. The stack of watered germination trays (maximum of 18) was wrapped in a transparent plastic bag to prevent moisture loss due to evaporation. The prepared stacks were placed in an incubator or climate chamber maintained at 25°C and continuous dark conditions. Germination was followed daily for up to 14 days by making images at frequent intervals using a digital camera (Nikon D80, http://www.nikon.com). Seeds were considered germinated if the radicle protruded by at least 2mm. Germination parameters, namely, total seed germination on the 14^th day (G_MAX, in percentage) and time for 50% germination (t₅₀, in hours) were calculated by automatically scoring the germination over time with GERMINATOR software (Joosen et al., 2010). Seedlings with an intact and healthy shoot and root system were evaluated as normal seedlings ( Supplementary Figure 1 ) as described in the ISTA Handbook on Seedling Evaluation (Don, 2006). Total normal seedlings (TNS; percentage) were determined on the 14^th day of the germination test.

To check whether the different aging treatments affected the metabolism of the rice seeds, metabolomic analysis of the lipid fraction and the volatile compounds was carried out, as detailed below.

The germination parameters used in this study were calculated by automatic scoring and curve fitting germination data in time (until 336hrs) for each genotype-treatment combination using the GERMINATOR package (Joosen et al., 2010). Germination parameters were calculated as the mean of the two replicates in the germination test. Descriptive statistics, statistical analysis, principal component analysis (PCA), and graphing were performed in OriginPro 2022 (https://www.originlab.com). Three-way analysis of variance (ANOVA) was applied to observe storage time-aging treatment, genotype-aging treatment-storage time and its interaction effects on seed germination parameters. Paired t-tests were used to compare the means of aging treatments at different storage time points. Pairwise Pearson’s correlation coefficient was estimated to compare seed germination and longevity (P ₅₀; see below) under different storage conditions. For all experiments, significance was determined as p<0.05 (*p<0.05, **p<0.01, ***p<0.001). Box plots represent standard box setting with first and third quartile split by the median and the whiskers extend to a maximum of 1.5× interquartile range beyond the box.

Probit analysis of seed survival data was performed using GenStat (19^th Edition, VSN International Ltd, United Kingdom, https://www.vsni.co.uk/software/genstat), to fit the seed viability equation (1) (Ellis and Roberts, 1980) and estimate seed longevity:

Where v is the probit proportion of viable seeds in a seed lot stored for a period of p days, K _i is the initial probit viability and σ is the length of time (days) for viability to fall by 1 probit. The P ₅₀ value, indicating the storage period required for viability to be reduced to 50%, was calculated by modeling the survival data according to Whitehouse et al. (2015). The higher the P ₅₀ value, the more tolerant the seeds are to aging treatment.

For those seed lots/accessions exhibiting dormancy, the breaking of dormancy is described by the equation:

in which nd is the probit proportion of non-dormant seeds in the seed lot after p days in experimental storage, K _nd is the fitted initial probit proportion of non-dormant seeds and σ_nd is the time it takes for the proportion of dormant seeds to reduce by 1 probit. Equations (1) and (2), were fitted simultaneously as a combined model describing the final observed germination (product of the proportions of non-dormant and viable seeds; Kebreab and Murdoch, 1999).

To establish an EPPO aging protocol for dry rice seeds, seeds of 20 different accessions were stored in a pilot experiment under ambient, EPPN, and EPPO conditions at 35°C for different storage durations. Under the control ambient pressure storage condition, a small aging effect was observed among the individual accessions, indicated by a decrease in total germination and normal seedlings and an increase in the germination rate (t₅₀) ( Figure 1 and Supplementary Table 3 ). After 56 DOS, seven lots showed a significant decline in total germination and an apparent delay in the rate of germination. Five of these lots also showed a significant decline in total normal seedlings. One seed lot (accession IRGC 117269) showed only 50% of normal seedlings after 56 DOS. Seed lots from three accessions (IRGC 117266, IRGC 117267, and IRGC 117283) initially showed some dormancy and after-ripening occurred during storage since their maximum germination values initially increased during the first 14 days, to remain at a high level for the subsequent days under ambient storage.

Under the pressure control (EPPN) storage condition, the average seed germination after 56 days (83%) was not significantly different from the ambient control (90%) ( Figure 1A and Supplementary Table 3A ). However, variation was observed in the response of individual seed lots. Out of the 20 seed lots analyzed, 10 showed a decline in maximum germination after 56 days storage under the EPPN treatment. These 10 included the seven seed lots that also showed a decline in total germination under ambient control storage. The seed lots of accessions IRGC 117273 and IRGC 117282 showed a specific sensitivity to the high-pressure storage, as their maximum germination was reduced after both 42 and 56 days under EPPN storage, while this reduction was not observed under ambient control storage. For 17 out of the 20 seed lots, storage under EPPN condition for 56 days resulted in a longer delay in seed germination rate (t₅₀) compared to the ambient control ( Supplementary Table 3C ). The three seed lots that showed dormancy and after-ripening under ambient storage showed the same under EPPN storage.

The EPPO treatment induced relatively faster deterioration of the seeds ( Figure 1 and Supplementary Table 3 ). Prolonged storage under EPPO resulted in a clear decline in average total germination, average normal seedlings, and delay in germination (higher t₅₀ values). Four seed lots (IRGC 117270, IRGC 117274, IRGC 117279, and IRGC 117282) were clearly more sensitive to EPPO storage than the others, as the maximum germination of these lots had declined below 10% after 28 days storage ( Supplementary Table 3A ). The average germination in three of these four seed lots declined relatively quickly under ambient control storage and for all four under EPPN storage. Two accessions (IRGC 117267 and IRGC 117283) were relatively more tolerant to the EPPO treatment, with more than 50% germinating seeds after 28 days. A similar pattern was observed for the percentage of normal seedlings ( Supplementary Table 3B ). While after the longest period (56 days) of storage under ambient or EPPN conditions, 80% or 74% of the total seeds, respectively, still developed into normal seedlings, under EPPO storage, this percentage rapidly declined from 51% at 14 DOS to 8% at 28 DOS and 0% at 56 DOS ( Figure 1B ). After 14 DOS, the seeds took longer to germinate (average t₅₀ of 113.6 h) when compared to the ambient control (63.5 h) and EPPN pressure control storage (69.5 h) ( Supplementary Table 3C ).

As the pressurized air used in the EPPO treatment is 78% nitrogen gas and a negative effect of EPPN, compared with the ambient control, was observed for a few seed lots, the germination data obtained under EPPO treatment were corrected for the seed lot-specific high-pressure sensitivity, depicted as ΔEPPO ( Supplementary Table 3 ). These pressure-corrected ΔEPPO values are already significantly different from the control treatments at 14 DOS for all the seed germination and seed quality parameters measured, indicating a specific and negative effect of elevated oxygen levels during seed storage.

The different seed germination and seedling quality parameters were highly correlated with each other, and the PCA clearly separated the results of the EPPO aging treatment and its corrected values (ΔEPPO) from the control treatments after 28 DOS ( Supplementary Figure 2 ). The first two principal components (PC1, PC2) cumulatively explained 93.8% ( Supplementary Figure 2B ) and 95.6% ( Supplementary Figure 2C ) of total phenotypic variation among rice accessions across different aging treatments at 14 and 28 days, respectively.

Germination data of the 20 rice seed lots following storage under EPPO as well as their pressure-corrected values (ΔEPPO) for different storage durations (up to 56 days) were fitted to estimate the seed longevity parameters of equation 1 (K _i, σ^-1) and P ₅₀, which showed clear differences among these lots ( Table 1 and Supplementary Figure 3 ). Under EPPO storage, after pressure effect correction (ΔEPPO), the initial seed viability (K _i) among the 20 rice seed lots ranged between 0.82 NED (IRGC 117274) to 3.10 NED (IRGC 117283) with a mean of 1.68 NED ( Table 1 ). The estimates of σ^-1 ranged between 0.041 (IRGC 117267) to 0.097 (IRGC 117283) with a mean of 0.073 NED d^-1. The P ₅₀ values also showed considerable variation among rice accessions when stored under EPPO: IRGC 117283 was the most tolerant with its longest P ₅₀ of 31.8 days, and IRGC 117274 was the most sensitive with a P ₅₀ of 12.0 days ( Table 1 ). The observed considerable genetic variation for sensitivity to elevated oxygen levels in this panel of 20 diverse rice accessions indicates that EPPO storage can be a useful experimental aging test to estimate the relative shelf-life of rice seed lots.

Since seed aging involves lipid oxidation processes, we studied the effects of EPPO treatment on the overall lipid composition in rice seeds from cv IR64 stored under EPPO and control conditions for different durations. The untargeted processing of the Liquid Chromatography-Mass Spectrometry (LC-MS) data resulted in the relative abundance values, corrected for the internal standard, of over 652 lipid-soluble compounds across all samples ( Supplementary Data 1 ). Differences in their lipid profiles between seed samples stored under different storage conditions are depicted in Figure 2 and Supplementary Figure 4 . A PCA, based on the relative abundance of 300 reproducibly detected compounds (using the criterium that the compound should be present in at least three samples at a signal abundance >15,000, i.e., three times the set detection threshold for peak picking) clearly separated the rice seed samples stored under EPPO from those of ambient control and EPPN control treatments ( Figure 2A ). The first two principal components explaining approximately 62% of the total variation (PC1 = 47.0%, PC2 = 14.7%), indicate that the 56 DOS EPPO-stored samples have a distinctive lipid profile compared to the ambient or 56 DOS EPPN stored samples. This is in agreement with the contrasting germination performance after the treatments. Specifically, the abundance of a series of lipid compounds eluting between 13.40 and 17.60 minutes was highly characteristic for the stronger-aged EPPO stored seed samples ( Figure 2B ). The results of a Hierarchical Cluster Analysis (HCA), based on the same 376 lipid compounds, are represented as a dendrogram and heatmap ( Supplementary Figure 5 ). The HCA confirmed that the profiles of the EPPO treated seeds have lipid profiles distinct from the two control treatments, with several lipid compounds specifically or much more substantially accumulating in the EPPO stored samples. There was no clear separate clustering of the two control treatments meaning that the EPPN treatment, which included a pressure induced ERH increase, did not result in apparent changes in the lipid profile compared to the storage at ambient pressure and ERH ( Supplementary Figure 5 ).

The metabolomics data were also explored to identify compounds correlating to seed viability and compounds accumulated during storage at each condition. Pearson’s correlations between the abundance of each lipid compound and the average percentage of seed germination of the respective samples were calculated ( Supplementary Data 1 ). The topmost compounds showing the highest negative correlations (r< -0.98) and highest positive correlations (r > 0.85) were selected for further annotation based on the observed accurate mass and potentially present molecular fragments ( Table 2 ). A negative correlation with seed viability, and thus positive with seed aging, was observed with compounds identified as oxidized lipids with one or more oxygen atoms incorporated ( Table 2 ). The only exception was Triacylglyceride (TAG) 52:3+O (m/z [M+Na]⁺ = 890.7812, cluster ID 483 in Supplementary Data 1 ) eluting at retention time 17.5 min), which was highest at the initial time point (0 day) and decreased to non-detectable levels upon subsequent EPPO storage. Apparently, this mono-oxygenated unsaturated TAG was already present in initial seeds and was further oxidized and converted into one of the TAGs of other lipid compounds increasing by EPPO storage and negatively correlating with germination performance. Lipid compounds showing highest positive correlation with seed viability, with a decline during seed EPPO aging, were identified as the lipophilic antioxidant coenzyme Q9 and unsaturated lipids typically found in rice seeds (e.g., tri-linoleoyl-glycerol, cluster ID 440 in Supplementary Data 1 ).

The untargeted approach to profile volatile compounds using HS-SPME-GC-MS identified 224 compounds emitted from ground rice seed samples after storage under different aging treatments. This number was reduced to 183 reproducibly detectable compounds after applying strict filtering criteria (using the criteria that the metabolites are present in all the replications of the QC samples) ( Supplementary Data 2 ). A PCA based on selected volatile compounds, where PC1 explains 33.2% of the total variation, clearly separates the samples aged under EPPO conditions from those aged under ambient and EPPN conditions ( Figure 3 ). An HCA performed on the same data confirmed the deviating volatile profiles of the EPPO treated rice samples ( Supplementary Figure 6 ). The heatmap showed the relative abundance of those volatile compounds accumulating over time, specifically in the EPPO stored samples.

Pearson’s correlation coefficients were calculated based on the relative abundance value of each volatile compound with percentage seed germination and storage duration ( Table 3 , Supplementary Data 2 ). Seeds stored under EPPO conditions were dominated by volatile compounds, which consistently and markedly increased as aging progressed, like 3,5-octadien-2-one, 2-methyl 2-propanol, hexanal, 2-heptanone, acetic acid, 2-butanone, heptanal, and 1-butanol.

To analyze to what extent dry EPPO aging mimics long-term natural aging, seeds of seven lots were also stored under conditions mimicking CWSS and TRSS. In addition, we compared these dry aging conditions with experimental aging at relatively high moisture conditions (CD test). Aging under both EPPO and CD treatment resulted in faster deterioration of the seeds, as shown in a decreased germination rate compared with aging under CWSS and TRSS conditions ( Supplementary Figure 7 ). Under CWSS and TRSS conditions, three seed lot (4256, 4258, and 4261) showed a negligible change in germination performance, even after prolonged storage. The rate of seed viability loss (estimated as the duration to 50% viability loss, P ₅₀) under the different storage conditions was calculated for each seed lot using the seed viability equation (equation 1 in the Materials and Methods section; Supplementary Table 4 ). Correlation analysis based on P ₅₀ values showed that ΔEPPO is significantly (p<0.05) positively correlated with both TRSS (r=0.89), CWSS (r=0.85) and CD (r=0.85) conditions ( Figure 4 ). The significant but relatively low correlation of 0.78 for P ₅₀ between EPPO and ΔEPPO indicates variation between seed lots in their sensitivity to high pressure (as deduced from the EPPN storage effects).

PCA was performed to compare the overall profiles of lipid-soluble compounds and headspace volatiles in rice seeds stored under the different aging conditions. The first two PCs together explained 47.4% and 54.5% of the total variation in the profiles for lipid compounds ( Supplementary Figure 8A ) and volatile compounds ( Supplementary Figure 8B ), respectively. Both PCA score plots did not show a clear separation of samples based on storage condition only, likely due to differential storage times imposing different levels of aging. A time-dependent effect is observed in these PCAs except for the samples aged under CD conditions. A more detailed comparison between the aging conditions was made for those samples closest to the storage time resulting in 50% viability (P ₅₀), i.e., 28d for EPPO, 9d for CD, 175d for TRSS, and 630d for CWSS. In total, 319 lipid compounds ( Figure 5A ) and 186 headspace volatile compounds ( Figure 5B ) were commonly detected across all four storage conditions at the P₅₀ time points. This considerable overlap may relate to the observed lack of storage-related clustering of samples in the unsupervised PCA ( Supplementary Figure 8 ). At their respective P ₅₀ effective time points, EPPO-stored seeds contained 21 unique lipid compounds, the CD contained 22, TRSS 18, and CWSS 32 lipid compounds ( Figure 5A ). Similarly, only two volatile compounds were unique to the CD storage and two others for CWSS ( Figure 5B ).

The metabolomics data were also explored to identify compounds correlating to the loss in seed viability and accumulation during storage at each aging condition ( Figure 6 ). A total of 89, 102, 94, and 79 lipid compounds in seeds stored under EPPO, TRSS, CD, or CWSS conditions, respectively, showed a high negative correlation (r > -0.50) with total percentage germination ( Supplementary Data 1 ), while for the volatile compounds these values were respectively 90, 89, 52, and 17 ( Supplementary Data 2 ). These correlating compounds were visualized by a Venn diagram to find unique and overlapping metabolites across all four storage conditions ( Figure 6 ). Interestingly, 24 lipid compounds and only one volatile compound were common to all four storage conditions, while 21, 14, 12, and 9 lipid compounds ( Figure 6A ) and 31, 43, 15, and 0 volatile compounds ( Figure 6B ) were found unique to EPPO, TRSS, CD, and CWSS conditions, respectively. Based on these correlating lipid compounds, EPPO storage had the highest overlap with both TRSS and CD (9 compounds each), followed by CWSS (6 compounds) ( Figure 6A ). Similarly, for the correlating volatile compounds, EPPO storage had the highest overlap with compounds found under TRSS (25), followed by CD storage (17) and CWSS (4) ( Figure 6B ).

The topmost lipid and volatile compounds showing the highest negative correlation with the percentage of seed germination were selected for further manual annotation based on their mass spectrometry data ( Supplementary Table 5 , 6 ). Most lipid compounds in seed samples stored under EPPO ( Supplementary Table 5A ) and CD conditions ( Supplementary Table 5D ) were identified as oxidized lipids with one or more oxygen atoms incorporated. Interestingly, lipids annotated in seeds stored under the two ambient long-term storage (CWSS & TRSS) showed a different lipid landscape compared to the two-fast aging assays ( Supplementary Table 5B, C ). Similarly, 3,5-octadien-2-one was among the major volatile compounds in seeds aged under EPPO conditions, which consistently increased many folds as aging progressed. 1-Butanol and acetic acid were found in seeds aged under EPPO, CWSS, and TRSS conditions but not in the CD samples ( Supplementary Tables 6A–C ). Major volatiles specific to CD storage included naphthalene and various benzene derivatives ( Supplementary Table 6D ).

The principle of experimental aging by EPPO is the acceleration of oxidation-induced damage through elevation of oxygen availability rather than aging by increasing moisture and temperature. A disadvantage of EPPO is that the increased pressure will theoretically increase the water activity (a_w) in seeds and that the build-up and decline in pressure, before and after the actual EPPO storage, can potentially induce physical damage to the seed. To correct for this physical pressure effect, the EPPN treatment was used as a control next to storage at ambient air pressure. With the EPPN treatment, the seed water activity will be similarly increased as with the EPPO treatment. Therefore, we expected that the EPPN treated seeds at 50% RH would age somewhat faster compared to ambient pressure stored seeds storage at 43% RH ( Figure 1 ). We observed that seeds from some accessions aged somewhat faster under the EPPN treatment compared to the ambient control, specifically regarding their slower germination (increase in t₅₀), while for other accessions there was no significant difference ( Supplementary Figure 3 ).

In contrast, there was no significant difference between ambient and EPPN controls for average total seed germination and seedling quality parameters ( Figure 1 ). Adverse effects of EPPN storage on seeds, resulting in a higher proportion of abnormal seedlings, have been reported for barley, with significant genetic variation for this sensitivity (Nagel et al., 2016). Since the observed negative effects of EPPN storage for some accessions depended on storage duration, it seems to be a direct or indirect physiological or biochemical effect during storage, rather than physical damage upon pressure build-up or pressure release at the start and end of EPPN storage, respectively. The theoretical slightly higher water activity induced by the higher pressure could be a contributing factor, but an effect of a high nitrogen pressure itself cannot be ruled out. An anesthetic effect by the increased partial pressure of nitrogen and some other gasses is well known with SCUBA diving, and their relative effect seems to be related to the gas lipid solubility (Bennett, 2004). The air used in the EPPO storage contains 78% nitrogen, resulting in a partial nitrogen pressure of 15.6 MPa, which approaches the 20 MPa in the EPPN treatment. Therefore, changes in germination parameters due to the increase in pressure were corrected in the EPPO treatment and represented as “ΔEPPO”, which is used to discuss the results observed in different experiments.

Our experiments with rice seeds under dry EPPO showed increased aging during storage, resulting in the typical symptoms of seed viability loss: slower germination and a decline in normal seedlings ( Figure 1 and Supplementary Table 3 ). It also confirms the negative effect of oxygen on rice seed shelf-life (Roberts, 1961), as was observed with other species (Justice and Bass, 1978; Schwember and Bradford, 2011; Groot et al., 2015). Our results are consistent with other studies using EPPO as an experimental dry aging method (Groot et al., 2012; Nagel et al., 2016; Buijs et al., 2020; Hourston et al., 2020; Renard et al., 2020) or older experiments on seed storage under increased oxygen pressure (Ehrenberg et al., 1957; Moutschen-Dahmen et al., 1959; Berg et al., 1965). Accelerated after-ripening under EPPO conditions, as previously reported for Arabidopsis (Buijs et al., 2018; Buijs et al., 2020), was also observed for some seed lots in our experiments ( Supplementary Figure 3 ). This dormancy loss was faster under EPPO storage compared to storage under ambient pressure or EPPN conditions, indicating a role of oxygen-dependent reactions in the after-ripening process.

Another important observation from our study is that the tested seed lots, representing different rice accessions, showed considerable variation in viability loss under EPPO storage ( Figure 1A ), indicated by the estimated P ₅₀ values ( Table 1 ). Although seed production and handling can influence subsequent shelf-life, we expect that the variation among the analyzed seed lots mainly reflects an interaction between the storage environment and genetics, as these seeds were produced at the same time and processed, handled, and stored similarly prior to our experiments. Large genetic plasticity for seed germination traits under EPPO aging was also found in barley (Nagel et al., 2016) and Arabidopsis (Buijs et al., 2020; Renard et al., 2020), where specific QTLs were identified. However, the analysis of genetic variation was not the aim of the experiments reported here.

Lipids are key molecules for energy storage, membrane functioning, and signaling. Lipid oxidation reactions induced by free radicals or ROS during storage are reported to be a primary cause of aging-related damage in seeds (Harman and Mattick, 1976; Johnson and Decker, 2015). In rice seeds, TAGs are the principal lipids (Morrison, 1995). These storage lipids are present in the form of lipid droplets or spherosomes in the aleurone layer, sub-aleurone layer, and embryo of rice seed (Juliano, 1983; Chapman and Ohlrogge, 2012). In the endosperm, these lipids are bound to protein bodies and starch granules. The fatty acid composition of rice seeds constitutes around 75-80% of unsaturated fatty acids, namely oleic acid (C18:1), linoleic acid (18:2), and linolenic acid (C18:3) (Tong et al., 2019). Considerable changes in lipid profiles, i.e. a decrease in the levels of both unsaturated fatty acids and phospholipids and an accumulation of lipid oxidation products like malondialdehyde (MDA), lipid peroxides, and oxidized TAGs, have been reported in response to rice seed aging (Gayen et al., 2014; Huang et al., 2014) and botanically related wheat seeds (Riewe et al., 2017). To obtain a comprehensive overview of changes in the seed lipidome, we performed an untargeted high-mass resolution LC-MS based lipidome comparison of dehulled rice seed samples stored for different durations. However, our goal was to identify only the main significantly altered lipid species rather than annotate as many detected compounds as possible. Different species from various lipid classes were represented in our rice seed samples, characterized by specific retention times and accurate masses of molecular ions ([M+H]+ or [M+NH3]+) ( Supplementary Data 1 ). These included TAGs, phospholipids, glycolipids, and free fatty acids and their oxidized forms. Evaluation of these lipidomics data using PCA and HCA could separate experimentally aged EPPO samples from less aged control samples ( Figure 2A , Supplementary Figure 5 ). The EPPO-aged seed samples had a relatively high abundance of oxidized lipid compounds ( Figure 2B and Table 2 ), some of which were previously reported to accumulate in long-term dry stored wheat seeds (Riewe et al., 2017). By analyzing seed samples taken at increasing time under EPPO storage, we showed that these oxidized lipids negatively correlate with the proportion of germinating seeds and positively correlate with storage duration ( Table 2 ). Due to the presence of unsaturated bonds in their molecular structures, polyunsaturated lipids are more susceptible to oxidative damage than saturated species. The oxidized lipids in our dry-aged EPPO seed samples are most likely the result of a non-enzymatic oxidation reaction (Oenel et al., 2017). However, enzymatic conversion by the action of lipid-soluble enzymes cannot be ruled out. For example, lipases that catalyze the hydrolysis of TAGs/diacylglycerides (DAGs), releasing fatty acids and glycerol, are stable and can still be active in dry wheat seeds even at low moisture levels (Barros et al., 2010; Wiebach et al., 2020). Because of the extremely low molecular mobility in dry seeds, such lipases should be close to their substrates. Although rice lipids will be exposed to lipases during seed grinding (Lam and Proctor, 2003), in our study, such potential lipase activity was prevented by grinding the seeds in liquid nitrogen and directly mixing the frozen powders with the organic extraction solvent. Interestingly, a single oxidized lipid species, i.e., TAG 52:3+O, showing the highest correlation with rice seed aging was also present in the initial seed material and thus might at least partly have been generated during seed processing and storage before the start of our experiments. Upon subsequent storage, such partially oxidized lipid may undergo further oxidation, e.g. into TAG 52:3+2O (m/z [M+Na]⁺= 906.7767 eluting at a retention time of 13.3 min).

The oxidative degradation of lipids was accompanied by the formation of lipid oxidation-related volatiles ( Table 3 , Supplementary Data 2 ), as has also been reported for seed aging in other crops (Mira et al., 2010). Also, with the volatile compounds as variables, the PCA discriminated seed samples stored under EPPO conditions from those stored under ambient or EPPN conditions ( Figure 3A ). A considerable number of lipid-derived volatile compounds accumulated more than 200-fold during EPPO storage and showed a strong negative correlation to loss in seed viability after the EPPO storage experiment ( Table 3 ). These correlating volatiles included compounds putatively identified as hexanal, heptanal, 2-heptenal, 2-heptanone, which are typical oxidation products of unsaturated fatty acids such as oleic acid, linoleic acid and linolenic acid (Wilkens and Lin, 1970; Frankel, 1983; Lam and Proctor, 2003). Similarly, several esters (for example, acetic acid-butyl ester that accumulated during EPPO storage) are among the oxidation-related degradation products from unsaturated fatty acids (Goicoechea and Guillén, 2014).

This untargeted lipid and volatile analyses proved that dry rice seeds stored under EPPO conditions (high oxygen levels) are sensitive to lipid oxidation reactions. These lipid oxidation reactions will also involve membrane lipids and are most likely the main reasons for viability loss. Additional oxidative reactions, not studied in the present experiments resulting in oxidative damage to DNA, RNA or proteins, can also be expected under EPPO conditions (Ohlrogge and Kernan, 1982; Farmer and Mueller, 2013). Further research is required to verify the oxidative stress-related changes in proteins or nucleic acids related to seed aging under EPPO storage conditions.

In general, longevity increases with a decrease in seed moisture content and temperature in a quantifiable and predictable way (Roberts, 1973; Ellis and Roberts, 1980). Seed moisture content, storage temperature, and chemical composition determine the physical state of the aqueous and lipid domain of the cytoplasm, and thereby the biophysical and biochemical processes related to seed deterioration (Vertucci and Roos, 1990; Ballesteros et al., 2020). Under dry conditions, the cytoplasm is in a glassy state with extreme low molecular mobility, while under humid conditions, the cytoplasm becomes liquid, increasing molecular mobility and allowing enzyme activity (Gerna et al., 2022). It is interesting to know if and how aging under EPPO storage may mimic aging under long-term natural conditions. However, ‘natural aging’ is not unambiguously defined, as it ranges between storage conditions that exist in highly controlled environments, like gene banks and commercial storage facilities for high value vegetable seeds, to rather uncontrolled environments such as in warehouses for relatively low value cereal seeds and open storage by farmers. In many studies, natural aging refers to laboratory bench storage for which storage conditions are less-well defined or not described at all. The RH in a laboratory can vary considerably over time. To perform a direct comparison, we stored rice seeds of seven different lots under four defined conditions. The conditions were EPPO storage (200 times higher oxygen levels, 35°C and theoretically at 50% RH), ambient pressure (air) conditions mimicking medium-term seed storage under dry conditions in controlled warehouses (CWSS, seed equilibrated to 40% RH and 28°C), short-term traditional storage practice (TRSS, seeds equilibrated to 60% RH and 28°C) and CD storage (seeds equilibrated to 75% RH and 35°C). The duration of storage in which the seed lot viability declines to 50% (P ₅₀) is considered a suitable measure of seed longevity (Hay et al., 2022a). Based on the P ₅₀ values ( Supplementary Table 4 ), it was observed that seeds from all lots lost their viability in shorter times under dry-EPPO storage or wet-CD storage compared to storage under the two ambient storage conditions (CWSS and TRSS). Results of correlation analysis for the P ₅₀ values for the seven seed lots indicated that high-pressure oxygen storage (ΔEPPO) showed significant (p<0.05) positive correlations with all the other storage conditions, while such significant correlations were not found between CD aging and either TRSS or CWSS nor between TRSS and CWSS ( Figure 4 ). It is worth noting that although the seed lots showed less variation for the slope of aging under EPPO storage, significant differences between seed lots were still detected ( Supplementary Table 4 and Supplementary Figure 7 ). Groot et al. (2012) showed that storage of dry lettuce seeds under EPPO conditions (at 20 MPa P_O2 and 40% RH) mimics dry aging symptoms as seen upon long-term storage in seed company warehouses (at ambient pressure and 30% RH) and that aging under EPPO conditions is physiologically different from aging under moist warm moist conditions at ambient pressure. They showed that in cabbage seeds, the level of tocopherol (vitamin E, a lipid-soluble antioxidant) linearly decreased with viability loss under EPPO storage but was more or less constant under CD storage at 85% ERH. A recent study using a diverse panel of Arabidopsis ecotypes showed that laboratory bench seed aging (40-60% RH, 20-25°C and 18 months) better correlates with a CD test (75%RH, 38°C & 14days; r=0.68) and EPPO (40% RH, 20°C & 5 months; r=0.47) compared to an accelerated aging test (100% RH, 39°C & 48 h; r=0.11) (Renard et al., 2020). In a QTL analysis with Arabidopsis, Buijs et al. (2020) compared EPPO storage at 20 MPa air after equilibration of the seeds at 35 or 55% RH (during the test likely 40 or 64% RH, when corrected for the theoretical pressure effect on water activity according to Okos et al. (2019)), with both CD-aging at 85% RH and lab bench-aging (undefined), and showed that the four storage conditions resulted partly in overlapping, partly in distinct QTLs. Also within these two EPPO treatments, both common and distinct QTLs were identified. Those results indicate that the response of the seeds is also influenced by the RH during the EPPO storage, as it is in other storage conditions.

In our EPPO treatment, the rice seeds were pre-equilibrated at 20°C and 40% RH, at which moisture level, we assume the seeds are still in a glassy state and metabolically inactive (Sun, 1997; Ballesteros et al., 2020; Hay et al., 2022b). In the EPPO system, aging occurs in closed chambers filled with high-pressure dry air and with silica gel pre-equilibrated at the same RH as the seeds to provide a buffer for any change in the seed moisture level due to filling with dry air (Bartolotta et al., 1991). According to Okos et al. (2019) the increased pressure (at 20 MPa) will, combined with the temperature effect on the water activity, give a theoretical increase in the water activity from 0.43 to 0.50, in accordance with 50% ERH. Unfortunately, our data loggers were not able to record the actual RH in the tank under pressure. Therefore, it still needs to be elucidated whether there is an actual and substantial increase in water activity during EPPO treatment. However, even when this is true, we presume that the cytoplasm in the embryo of rice seed at 50% RH and 35°C is not in the liquid state. This is supported by the observation of an apparent effect of increased oxygen levels on seed aging, which is not expected for seeds with a liquid cytoplasm (Gerna et al., 2022).

In this study, we also analyzed the biochemical changes in the seeds, as deduced from untargeted lipidomics and headspace volatile analyses. We observed that seed samples at the start of storage (0 DOS) for each of the four storage conditions did not cluster together. This observation may indicate already some metabolic changes induced during the preparation of the samples for the different storage treatments, which is mainly a temporary change in seed water activity for a few days. Upon subsequent storage-induced and aging to a more or less similar germination percentage (P ₅₀), we observed a large overlap in the presence of lipid species (319 out of the 652 compounds in total) and headspace volatiles (186 out of the 224) between all four storage conditions ( Figure 5 ), indicating that many of the metabolic changes are common for all four aging conditions. These metabolites are mainly concerned with oxidized lipids. A more detailed analysis of compounds correlating (r< -0.50) in their relative abundance with loss in seed viability indicated that EPPO samples showed the highest overlap with samples aged under TRSS, followed by CWSS and CD storage ( Figure 6 ). However, there was hardly an overlap between the most humid CD storage and the least humid CWSS storage for either lipids or volatiles, indicating apparent differences between both storage regimes in inducing biochemical changes in the seed ( Figure 6 ). Chemical identification of the topmost correlating lipids indicated an increase in oxidized TAGs in both EPPO and CD samples compared to seeds stored under the two ambient storage conditions ( Supplementary Table 5 ). The possible reason could be the warmer conditions used in the experimental aging conditions (35°C) compared to the temperature at TRSS and CWSS (28°C). The accumulation of oxidized lipids coincides with the reduction of unsaturated TAGs, most prominently in the EPPO-aged samples. It is important to note here that in TRSS and CWSS samples, the accumulation of oxidized lipids involved a lower number of different TAG species ( Supplementary Table 5B, C ). Increased levels of oxidized lipids have also been observed with wheat seeds stored at around 6% RH and 20°C temperature conditions compared to seeds stored at 13% RH at 0°C, followed by -18°C conditions (Riewe et al., 2017; Wiebach et al., 2020). CD-aged samples contained more oxidized DAGs in addition to TAGs, presumably due to the hydrolytic cleavage of (oxy)PLs (Riewe et al., 2017).

Relative levels of 1-butanol (a volatile compound identified in oxidized soybean oil (Frankel, 1983)) were increasing while germination rates decreased in both EPPO, CWSS, and TRSS samples, but not in CD samples ( Supplementary Table 6 ). Similarly, acetic acid, a product of the Maillard reaction (Mancilla-Margalli and López, 2002), also consistently increased with loss in seed viability in EPPO, CWSS, and TRSS samples, but it showed an opposite trend in CD samples (r=0.85). Hexanal, a secondary product of linoleic acid oxidation (Lam and Proctor, 2003; Colville et al., 2012), was only found to increase as aging progressed in EPPO samples. The naphthalene and benzene derivatives observed among the volatiles from seeds stored under CD conditions were unexpected. They might be considered as contamination from the CD test performed in laminated foil pouches. In our study, both the lipidomics and headspace volatile analyses showed common and distinct biochemical changes in seeds aged under ambient (CWSS and TRSS) conditions ( Figure 6 ). Taking the metabolite analyses together, under all four storage conditions tested, many of the accumulating metabolites are common and mainly lipid oxidation products, showing oxidation as a main deteriorating effect during seed aging. The correlation analysis showed that regarding the lipids or volatile metabolite pattern, EPPO storage at 50% ERH resembles better to TRSS at 60% compared to CWSS at 40% ERH and CD at 75% ERH, indicating a clear effect of ERH in a more detailed look on the different lipid oxidation products that are accumulating. The latter will be a result of production and downstream conversion.