Introduction
Seed longevity (i.e. the duration seeds survive during storage) is fundamental to seed companies and plant genebanking operations (Hay and Whitehouse, Reference Hay and Whitehouse2017). We typically think of seed longevity as the threshold between when seeds retain and lose the capacity to germinate (Rajjou and Debeaujon, Reference Rajjou and Debeaujon2008; Walters et al., Reference Walters, Ballesteros and Vertucci2010; Hay et al., Reference Hay, Davies, Dickie, Merritt and Wolkis2022). This threshold of functionality is analogous to ‘expiration dates’ for other materials such as dried foods and pharmaceuticals, and can vary considerably across species and seed lots within a species (Walters et al., Reference Walters, Wheeler and Grotenhuis2005, Reference Walters, Ballesteros and Vertucci2010; Nagel et al., Reference Nagel, Vogel, Landjeva, Buck-Sorlin, Lohwasser, Scholz and Börner2009; Probert et al., Reference Probert, Daws and Hay2009; Lee et al., Reference Lee, Jeon, Lee, Lee and Kim2013; Willis, Reference Willis2017; Pritchard, Reference Pritchard2020). Predicting expiration dates for seeds remains elusive because numerous interacting factors are involved and underlying mechanisms contributing to each factor are poorly understood.
Longevity is often characterized by a cut-off value marking changed functionality, such as P85 or P50, the time at which the sample is 85 or 50% viable, respectively. Plant genebanks use an 85% viability metric because this appears to minimize risk of genetic erosion as well as ensure successful stand establishment during regeneration (FAO, 2014). Because P85 cannot be predicted, genebanks rely on periodic testing of stored seeds, which provides ‘snap-shots’ of seed germination potential. This does not mitigate the issue of predicting longevity, because the frequency of testing heavily relies on crude estimates of seed longevity. Testing intervals of 10–20 years are recommended for genebanks storing seeds at −20°C. Thus, a sample that survives 100 years may be tested 5–10 times, consuming hundreds to thousands of valuable seeds. Too frequent or infrequent testing poses major risks of loss within genetic resource collections, either by depleting long-lived samples or samples dying in between monitor tests (Hay et al., Reference Hay, de Guzman, Ellis, Makahiya, Borromeo and Hamilton2013; Fu et al., Reference Fu, Ahmed and Diederichsen2015; Hay, Reference Hay and Dulloo2021). Tools are needed to reliably predict longevity (the time that a seed remains viable) or to measure ageing rate (i.e. longevity−1) at times before P85. Unfortunately, ageing in seeds is mostly characterized by the final effect of ageing – mortality – rather than its less catastrophic symptoms. Because there are no recognized symptoms of ageing besides germination behaviour, it is a challenge to assess the onset of viability loss in a viable seed lot.
Seed longevity, and its reciprocal, seed ageing, are dependent on the storage conditions (moisture and temperature) and intrinsic properties of the seed collectively referred to as ‘initial quality’. Moisture and temperature control the molecular mobility within the cytoplasm of seeds and are major contributors to ageing kinetics (Ellis, Reference Ellis2022; Nadarajan et al., Reference Nadarajan, Walters, Pritchard, Ballesteros and Colville2023; Rao et al., Reference Rao, Pallavi, Bharathi, Priya, Sujatha and Prabhavathi2023). These principles are increasingly understood in the context of material sciences related to glassy properties (Ballesteros and Walters, Reference Ballesteros and Walters2019). Given moisture and temperature conditions, seed longevity can be approximated from empirically based models (Ellis, Reference Ellis1988; SER, 2023) within boundaries of about 20–25% relative humidity (RH) and −30 to −40°C (Vertucci and Leopold, Reference Vertucci and Leopold1987; Ellis et al., Reference Ellis1988; Ellis and Hong, Reference Ellis and Hong2006).
The components of initial seed quality that contribute to seed longevity remain elusive, mostly because they are both hard to quantify and to control. Typically, species may be characterized as producing short-lived (e.g. onion) or long-lived (e.g. oat) seeds and numerous studies report relative longevity among species or survival times expected in warehouse or genebank storage (Nagel and Börner, Reference Nagel and Börner2010; Thirusendura Selvi and Saraswathy, Reference Thirusendura Selvi and Saraswathy2018; Solberg et al., Reference Solberg, Yndgaard, Andreasen, Von Bothmer, Loskutov and Asdal2020). Importantly, genebanks report high, unexplained variation of longevity among accessions within a species despite similar provenances (Walters et al., Reference Walters, Wheeler and Grotenhuis2005; Nagel et al., Reference Nagel, Vogel, Landjeva, Buck-Sorlin, Lohwasser, Scholz and Börner2009, Reference Nagel, Rosenhauer, Willner, Snowdon, Friedt and Börner2011; Solberg et al., Reference Solberg, Yndgaard, Andreasen, Von Bothmer, Loskutov and Asdal2020). Studies of ecotypes and QTLs regulating longevity of dry seeds strongly suggest genetic mechanisms (Kueneman, Reference Kueneman1983; Bentsink et al., Reference Bentsink, Alonso-Blanco, Vreugdenhil, Tesnier, Groot and Koornneef2000, Reference Bentsink, Hanson, Hanhart, Blankestijn-de Vries, Coltrane, Keizer, El-Lithy, Alonso-Blanco, de Andrés and Reymond2010; Clerkx et al., Reference Clerkx, El-Lithy, Vierling, Ruys, Blankestijn-De Vries, Groot, Vreugdenhil and Koornneef2004; Schwember and Bradford, Reference Schwember and Bradford2010; Nagel et al., Reference Nagel, Rosenhauer, Willner, Snowdon, Friedt and Börner2011; Rehman Arif et al., Reference Rehman Arif, Nagel, Neumann, Kobiljski, Lohwasser and Börner2012; Agacka-Mołdoch et al., Reference Agacka-Mołdoch, Nagel, Doroszewska, Lewis and Börner2015; Zhang et al., Reference Zhang, Hina, Song, Kong, Bhat and Zhao2019). This is consistent with numerous studies of genetic control of desiccation tolerance in diverse organisms (Hesgrove et al., Reference Hesgrove, Nguyen, Biswas, Childs, Shraddha, Medina, Alvarado, Yu, Sukenik, Malferrari, Francia, Venturoli, Martin, Holehouse and Boothby2021; Nguyen et al., Reference Nguyen, Kc, Gonzalez, Tapia and Boothby2022; Packebush et al., Reference Packebush, Sanchez-Martinez, Biswas, Shraddha, Nguyen and Boothby2022). Maturity at harvest and pre- and post-harvest treatments also influence seed survival during storage (Argerich and Bradford, Reference Argerich and Bradford1989; Bradford et al., Reference Bradford, Steiner and Trawatha1990; Grass, Reference Grass1994; Hay and Probert, Reference Hay and Probert1995; Pinheiro et al., Reference Pinheiro, Dias, Medeiros, Ribeiro, Silva and Silva2021), clearly indicating that longevity is a complex trait influenced by genetic × environmental (G × E) interactions.
A correlation between initial germination percentage and performance during storage is often assumed (e.g., Ellis, Reference Ellis1988). Logically, progressive deviation from 100% viability places the seed lot in ever-closer proximity to the longevity threshold when seeds begin to die rapidly and this would indicate that ageing is occurring. However, tracking aging by subtle changes in germination requires large sample sizes and frequent monitoring. Also, some seed quality traits that affect initial germination are neutral to seed longevity, and some highly vigorous seed lots age rapidly (e.g. willow (Ballesteros and Pence, Reference Ballesteros and Pence2017)). An evolving concept of ageing-induced mortality uses ‘the straw that breaks the camel's back’ analogy, whereby a minor event can have a major effect. In this concept, germination and longevity are disassociated and accumulation of minor events (i.e. ‘straws’) will have mostly undetectable effects on viability. In other words, ageing is not detectable by viability loss until it is too late and the longevity threshold is reached without warning.
Stress tests can help unify various concepts of ageing mechanisms by addressing how much stress can be endured before a seed lot succumbs (Silva et al., Reference Silva, Soares, Martins, Silva, Silva and Silva2024). This information may provide a better starting point than % germination for empirical models of longevity. One common approach named ‘accelerated ageing’ or ‘controlled deterioration’ places seeds in unfavourable conditions, usually a combination of high humidity (≥75% RH) and high temperature (>45°C). Seed lots that succumb first are expected to die soonest even under more favourable conditions, that is dry (10–25% RH) and cold temperatures (−18 to 5°C). These kinds of tests come under perennial criticism because they are hard to translate back to actual genebanking conditions. Genebank conditions approach, or extend beyond, boundary conditions of empirical models, with limited data to verify that seed lots identified as relatively short- and long-lived in a stress test perform as expected in the genebank.
An alternative approach is to seek assays that detect degradation within stored seeds before the longevity threshold – that is, an assay of non-lethal ageing rather than a viability assay. Linking the kinetics of presumed reactions of dried materials with longevity (or longevity−1) involves waiting for the longevity signal; the waiting period could be decades to confirm that the assay is relevant to the dry and cold conditions of genebanks. One candidate assay involves assessing RNA fragmentation, which was studied in legacy collections that were 30 and 60 years old (Fleming et al., Reference Fleming, Hill and Walters2019). RNA molecules are labile and are rapidly degraded by RNases in actively metabolizing cells (Wurtmann and Wolin, Reference Wurtmann and Wolin2009). However, in dry seeds, RNase activity is low and RNA molecules persist during storage. RNA molecules tend to fragment at random which can be quantified from total cellular RNA using RNA integrity number (RIN) (Fu et al., Reference Fu, Ahmed and Diederichsen2015; Fleming et al., Reference Fleming, Richards and Walters2017, Reference Fleming, Hill and Walters2018a, Reference Fleming, Patterson, Reeves, Richards, Gaines and Walters2018b; Walters et al., Reference Walters, Fleming, Hill, Dorr and Richards2020; Tetreault et al., Reference Tetreault, Fleming, Hill, Dorr, Yeater, Richards and Walters2022). Initial studies show that RIN decline occurs linearly and this decline is detected in all tested species before reaching the longevity threshold that marks rapid mortality (Fleming et al., Reference Fleming, Hill and Walters2019). This assay seems especially promising because the temperature dependency of RIN decline in soybean and other species was indistinguishable from temperature dependency for viability loss, which indicates that factors controlling molecular mobility affect both processes in similar ways (Fleming et al., Reference Fleming, Hill and Walters2019; Walters et al., Reference Walters, Fleming, Hill, Dorr and Richards2020).
To be a valid indicator of ageing rate (longevity−1), the rate of RIN decline should correlate with factors in addition to temperature that affect seed longevity, namely, moisture and seed quality. The work presented in this paper uses a 20-year-old legacy collection of different varieties of chickpea, pea and lentil with similar harvest and storage provenances (Redden and Partington, Reference Redden and Partington2019). The original purpose of the collection was to demonstrate variation of seed longevity within a species that was independent of initial germination percentages (the standard method to quantify initial seed quality) (Redden and Partington, Reference Redden and Partington2019). Here, we hypothesized that RIN values would be higher in longer living species and longer living cultivars within a species. Taking advantage of the large seed size of legumes, we extended the study to investigate RIN decline at the tissue level and hypothesized that RIN decline would be comparable in seed tissue from cotyledons and embryonic axes. Experiments comparing ageing of varieties used 20°C storage and a single moisture range. Additional samples were stored at 2°C and different moisture levels, allowing us to test correlations of RIN with storage moisture and confirm temperature effects reported previously in an independently designed experiment.
Materials and methods
Seed treatment design
Seeds from ten lines of chickpea (Cicer arietnum L.), pea (Pisum sativum) and lentil (Lens culinaris Medikus subsp. culinaris) were used in this study. All accessions were grown in 2002 from regeneration plots at the Australian Grains Genebank (AGG, Horsham, Victoria, Australia) as described previously (Redden and Partington, Reference Redden and Partington2019). Water content of the seeds was adjusted as previously described (Redden and Partington, Reference Redden and Partington2019), and the seeds were sealed in three-ply laminate pouches with an aluminium foil layer that was at least 15 μm. Each pouch represented a sampling time and pouches were placed at either 2 or 20°C (Table 1). Seeds stored at 2 and 20°C represent ‘Part A’ and ‘Part B’, respectively, of a larger experiment described in Redden and Partington (Reference Redden and Partington2019). In Part A, water content of two lines of chickpea, pea and lentil were adjusted to 9–12% (high), 7–10% (medium) and 5–8% (low). In Part B, water content of seeds from ten lines of each crop was adjusted to 7–10% (medium).
Table 1. Environmental treatments of seeds used in this study
Water contents (wc) were adjusted in a bulk sample of each genetic line and temp × wc treatment in 2003-harvested seeds. Seeds were then sealed in foil laminate bags and stored at the Australian Grains Genebank (AGG) at either 20 (9–10 lines) or 2°C (2 lines). Water content was resampled for this experiment in 2016 and 2019 (20°C, AGG) and 2023 (20 and 2°C, NLGRP), and was expressed here as a percentage (%) on a fresh weight basis. Cell entries of ‘no data’ indicate that water content was not measured to conserve seeds for germination testing. Empty cell entries indicate that this temp × wc treatment did not occur. To provide an additional assessment of moisture effects, an independent dataset of pea seeds stored for 7 years at various RHs at room temperature (20–22°C) is also provided. A range of RH is provided by direct measurement using a BlueMaestro wireless Tempo Disc™.
A subset of the seeds was sent to the National Laboratory for Genetic Resource Preservation (NLGRP, Fort Collins, Colorado) in 2016 for RIN analysis and in 2023 for RIN, water content and germination measurements. Seed germination and water content were measured upon receipt at NLGRP. Seeds were stored at −20°C until RIN measurements were completed.
An additional study of the effect of moisture on RIN decline was conducted at NLGRP using pea cv. ‘Alaska’ seeds stored from 2017 to 2023 at room temperature and different RHs, obtained using saturated solutions of LiCl (13% RH), MgCl2 (33% RH), K2CO3 (43% RH) and CaNO3 (52% RH). These represent the same lot used for the initial time point in Fleming et al. (Reference Fleming, Hill and Walters2019) in which initial germination was 99% and the initial RIN values were reported as 7.8 ± 0.3 (axes) and 8.0 ± 0.7 (cotyledons).
Seed moisture
Water content of seeds was initially measured at AGG in 2003 using a Mettler LP16/PM400 seed moisture meter (Redden and Partington, Reference Redden and Partington2019) and subsequently for this paper in 2016 and 2019 by the high-constant-temperature oven method (ISTA, 2003) (Table 1). Water content was additionally measured at NLGRP in 2023 by comparing fresh and dry mass of seeds, dry mass determined by placing seeds at 95°C for 3 d (Table 1). There was some variation in water content measurements among pouches of the same treatment which could indicate experimental uncertainty or water movement over time if pouches were not perfectly waterproof.
RH surrounding the seeds can be estimated from water content measurements using generic isotherms constructed for species (SER, 2023). RH was measured at room temperature in 2023 (NLGRP) by placing a wireless RH sensor (BlueMaestro, London, UK) into the foil pouches containing seeds immediately upon opening them. Measured (Table 1) and SID RH values were consistent. The ‘low’ moisture treatment corresponded to 14–27% RH for all species. There was greater variation of RH at ‘medium’ and ‘high’ treatments ranging from 25 to 47% RH (pea and lentil at medium), 38 to 51% RH (chickpea at medium), 38 to 58% RH (pea and lentil at high) and 55 to 67% RH (chickpea at high).
The water contents of pea seeds cv. Alaska in precisely controlled RH chambers were comparable to the measured RH as well as the SID isotherm models (Table 1, bottom).
Germination assays and determination of longevity
Change in seed quality over storage time was measured using an initial germination assay and subsequent assays were conducted almost yearly at first and then spread out at 4- to 6-year intervals. Methods used at AGG (before 2023 assay) followed the International Seed Testing Handbook (ISTA, 2003); 100–200 seeds were spread over a paper towel, placed at 20°C and normal seedling development was assessed at 5 and 7 d (Rao et al., Reference Rao, Hanson, Dulloo, Ghosh, Nowell and Larinde2006; Redden and Partington, Reference Redden and Partington2019). Similar methods were used in the 2023 assay conducted at NLGRP except that 50 seeds were used for germination assays.
The longevity of seed lots was calculated by fitting time course germination data across the 20 years of storage to an Avrami equation (Avrami, Reference Avrami1941; Walters et al., Reference Walters, Wheeler and Grotenhuis2005) to determine P50. The value of P50 (storage time to which germination declines to 50%) was interpolated or extrapolated (for long-lived samples in which germination was >50% at the time of publication). The ageing rate is expressed as P50−1.
The expected deterioration of each species was also fitted to the Ellis and Roberts longevity model as posted on the Seed Information Database (SER, 2023, visited 20 August 2023) using initial viability = 99%; storage temperature = 20°C; water content = 7.7, 9.9 and 9.6% (chickpea, pea and lentil) and species constants (Ellis, Reference Ellis1988; Ellis et al., Reference Ellis1988; Whitehouse and Norton, Reference Whitehouse and Norton2022).
RNA degradation (RIN) assays
RNA degradation was quantified using RIN of 4–12 individual seeds per line separated into embryonic axes and cotyledons. RIN assays were measured in 2017 and 2023.
RNA extractions were used between 1 and 20 mg of dry tissue. Tissue was ground, in the presence of liquid nitrogen and 1–2 mg polyvinylpyrrolidone-40 (PVP-40; Fisher Scientific, Fair Lawn, NJ), in a 2 ml centrifuge tube using a nickel/lead steel shot bead (Ballistic, Inc., Hamel, MN) using the TissueLyser II (Qiagen, Hilden, Germany). RNA was extracted from ground tissue using the Qiagen Plant RNeasy kit (Qiagen, Hilden, Germany) following the manufacturer's instructions. RNA yield was quantified using a DeNovix DS-11 FX+ Spectrophotometer (DeNovix, Wilmington, DE). Samples were diluted to 2 ng μl−1 in nuclease-free water.
The RIN of diluted RNA samples was quantified on an Agilent Bioanalyzer (Agilent, Waldbronn, Germany), using Agilent RNA 6000 Pico chips and the Plant RNA Pico assay (Agilent 2100 Expert software version B.0208.SI648 R3), following manufacturer protocols. Fragment sizes of diluted RNA were assessed by electrophoresis using Agilent RNA 6000 Pico chips followed by analysis with the Agilent 2100 Expert software where electropherograms are assessed for peak areas of different fragment sizes and assigned a RIN value using a proprietary formula that involves peak sizes for rRNA (Fleming et al., Reference Fleming, Richards and Walters2017; Tetreault et al., Reference Tetreault, Fleming, Hill, Dorr, Yeater, Richards and Walters2022).
Statistical and data analysis
Linear regressions of RIN and most recent germination proportion, P50 and P50−1, were calculated using Excel ‘linest’ functions. Linear regressions were first performed using data from all three species combined to represent effects of species having different seed longevities. Subsequent linear regressions were performed for each species separately to examine the within-species variation of aging rates and RIN decline. To estimate the confidence of a P50 estimate for the 30 varieties, germination data were also modelled using a general linear model with binomial error distributions following the ‘dose.p’ function using the statistical environment R (R Core Team, 2019). Analyses of variance tests were performed within species to determine significance in the RIN values from seeds stored at 2°C with varying moisture contents; post-hoc Tukey's honestly significant difference (HSD) tests were conducted to assess the significant differences between RIN means. Analyses of variance tests were performed using JMP 12.2.0 (JMP Statistical Discovery LLC 2022).
Results
Changes in seed germination during storage
Initial seed germination was high for all varieties studied, being greater than 90% and mostly between 98 and 100% (Tables 2 and 3). A measurable effect of moisture environment on seed longevity was difficult to discern for seeds stored at 2°C even after 20 years of storage (Table 2). Germination remained relatively high for most treatments except the high moisture treatment of chickpea cv. ‘Howzat’ (56% germination) and the low moisture treatment of pea cv. ‘Kaspa’ (83% germination) (Table 2). The effect of moisture environment was apparent in a subsidiary experiment using pea seeds stored at room temperature (20–22°C) for 7 years. In this case, germination declined substantially in seeds stored at RH higher and lower than 33% RH. Maximum germination of 90% (for pea seeds stored at 33% RH) was slightly reduced from initial values (Fleming et al., Reference Fleming, Hill and Walters2019; Table 2). The lower germination of seeds stored at 13 compared with 33% RH was previously reported (Vertucci and Roos, Reference Vertucci and Roos1990).
Table 2. Effect of moisture during storage on germination of stored seeds
In the top portion of the table, seeds from two genetic lines per species were harvested in 2003 and stored at 2°C, indicating moisture levels (Table 1). The lower portion of the table shows a comparable experiment with seeds harvested in 2015 and stored at controlled RH on the lab bench (room temperature) for 7 years. Germination assays were conducted in 2023.
Table 3. Change in the germination of 9–10 genetic lines of chickpea, pea and lentil that were stored for 20 years at 20°C and 7–10% water (Table 1, 29–47% RH)
Germination was tested periodically, and some samples were depleted by 2016. Germination data (here and Redden and Partington, Reference Redden and Partington2019) were fitted to the Avrami equation. In several cases, the model was extrapolated past 50% germination in order to calculate the longevity (P50, storage time to 50% germination) of each genetic line.
Germination declined in most of the AGG seeds that were stored at 20°C and 30–49% RH (Table 3). Early changes in viability were reported previously (Redden and Partington, Reference Redden and Partington2019) and new data are presented here for assays conducted in 2016, 2019 and 2023 (Table 3). Average germination percentage for the accessions tested in 2023 was 89, 54 and 36% for chickpea, pea and lentil, respectively. However, there was large variation in germination among cultivars, with final germination ranging from >90 to <10% in all species, despite similar provenance, initial quality, storage time and storage environment. High germination (>90% in 2023) was noted in seven out of ten chickpea lines, one pea (‘Parafield’) and one lentil line (‘Cumra’). Hard-seededness may play a role in the slower aging of pea cv. Parafield seeds, which may have been stored in a lower moisture environment (Redden and Partington, Reference Redden and Partington2019).
To calculate a seed longevity parameter, data for germination % versus time at 20°C storage were fit to the Avrami equation for each AGG cultivar (Fig. 1) and time to 50% germination (P50) was determined. P50 ranged considerably within species: 11–68 years for chickpea, 13–41 years for pea and 15–20 years for lentil (Table 3). Average P50 for the species was 36, 18 and 16 years for chickpea, pea and lentil, respectively, suggesting that chickpea seeds were longer lived than pea or lentil. Regression analyses between final germination and P50 calculated using the Avrami model were somewhat weak when all species were combined: R 2 = 0.34, P < 0.0001, P50 = 11.6 + 0.17 × FinalGerm; for pea: R 2 = 0.47, P = 0.016, P50 = 11.4 + 0.16 × FinalGerm; for lentil: R 2 = 0.49, P = 0.014, P50 = 14.4 + 0.10 × FinalGerm, the relationship is not statistically significant for chickpea. Relationships between final germination and P50−1 were stronger due to lower uncertainty when longevity is expressed as ageing rate in samples that exhibited minor deterioration within the 20-year time frame (all species combined: R 2 = 0.57, P > 0.0001, P50−1 = 0.07–0.00036 × FinalGerm; for chickpea: R 2 = 0.65, P = 0.002, P50 = 0.095–0.00058 × FinalGerm; for pea: R 2 = 0.62, P = 0.003, P50 = 0.075–0.00037 × FinalGerm; for lentil: R 2 = 0.47, P = 0.017, P50 = 0.072–0.00044 × FinalGerm).
Fig. 1. Germination data and viability curves for varieties of (A) chickpea, (B) pea and (C) lentil at 20°C for 20 years with water content ranging from 0.049 to 0.11 g H2O g−1 dm. Germination data for each variety measured over the 20 years of the experiment are fitted using the Avrami method.
To test the consistency of seed longevity observed here with modelled longevity representative of the species, we calculated deterioration time courses from the viability equation (SER, 2023) using species constants (Ellis, Reference Ellis1988; Ellis et al., Reference Ellis1988; Whitehouse and Norton, Reference Whitehouse and Norton2022) and initial germination = 99%, temperature = 20°C, water content = 7.7, 9.9 and 9.6% for chickpea, pea and lentil, respectively. Viability equation models yielded P50 estimates of 29 years for chickpea and pea and 18 years for lentil, values which are relatively consistent with average P50 values calculated from Avrami curve-fitting for chickpea and lentil. No within-species variation is predicted from the viability equations given similar starting germination percentages and species constants for temperature and moisture.
RIN of stored seeds
RIN was measured in seeds stored at 2°C at different moisture levels, pea seeds stored at room temperature under precisely controlled RH, and a representative selection of cultivars stored at 20°C and 30–40% RH. RIN was assayed from seeds removed from storage in 2016 (AGG, 20°C storage only) and in 2023. RNA was extracted from individual seeds and cotyledon and embryonic axis tissues separately.
Quality of RNA extracted from seeds stored for 20 years at 2°C was excellent, with distinguishable rRNA peaks (slowly eluting region) and little evidence of short RNA fragments in fast eluting regions of the electropherogram (Supplementary Fig. S1). RIN values for the medium and low moisture treatments were not significantly different in most cases and ranged from 8.6 (pea axes) to 8.1 (chickpea axes and cotyledons) (Table 4). In contrast, RIN values for the high moisture treatment were usually lower and ranged from 7.7 (lentil cotyledon) to 7.1 (chickpea axes or cotyledons).
Table 4. Effect of moisture on quality of RNA extracted from long-term stored seeds
RNA quality is expressed as the RIN. In the top portion of the table, seeds from two genetic lines per species were stored at 2°C and indicated moisture levels (Table 1) between 2003 and 2023. The lower portion of the table shows a comparable experiment with seeds stored at controlled RH on the lab bench (room temperature) for 7 years and the comparable 2016 Alaska pea seed stored at 5°C. Superscript letters indicate a significant difference among treatments for the genetic line according to the post-hoc Tukey's HSD test.
A trend of decreasing RIN values with increasing storage moisture was observed in pea seeds stored at room temperature in which RIN decreased from 7.6 to 5.0 in axes stored at 33–52% RH for 7 years. RIN of pea seeds harvested in 2015 and measured in 2017 was 7.8 (Fleming et al., Reference Fleming, Hill and Walters2019) and so comparable to those of seeds stored at 2°C (RIN = 7.8 in 2023; Table 4).
A subset of the AGG seeds stored at 20°C was removed from storage in 2016 and in 2023. Average RIN values for grains removed in 2016 were 7.7 (±0.1 SE, n = 44), 5.2 (±0.2 SE, n = 81) and 5.9 (±0.30 SE, n = 65) for chickpea, pea and lentil, respectively (Table 5). Average RIN decreased to 7.4 (±0.10 SE, n = 40), 4.8 (±0.30 SE, n = 68) and 5.3 (±0.40 SE, n = 76) for chickpea, pea and lentil, respectively, when sampled in 2023, a difference of 0.3 (P < 0.0001), 0.4 (P = 0.3612) and 0.6 (P = 0.0299), respectively, over 7 years (Table 5).
Table 5. The quality of RNA extracted from different cultivars of legume seeds stored at 20°C and indicated moisture (Table 1) since 2003
RNA quality is expressed as the RIN. Seeds were removed from storage at 20°C in 2016 and 2023 and placed at −20°C until RNA was extracted (some in 2017, but mostly in 2023).
The average RIN values from embryonic axis or cotyledon tissue from the same seed were significantly correlated for chickpea, pea and lentil (P = 0.0017, P < 0.0001 and P < 0.0001, respectively; Fig. 2). Although correlated, the values were not identical. Average RIN values tended to be slightly greater for embryonic axes compared with cotyledons in undegraded seeds (germination > 90%; RIN > 8.0) and then lower in embryonic axes compared with cotyledons as seeds showed symptoms of ageing (germination < 80%; RIN < 7.0, with the exception of pea cv. Parafield). This tendency is noted as points along the 1:1 line at high RIN and below the 1:1 line at low RIN (Fig. 2). This result of slightly greater RIN degradation in embryonic tissue compared with cotyledon tissue was not significantly different within species (P = 0.6804, P = 0.3805 and P = 0.0850, chickpea, lentil and pea, respectively).
Fig. 2. RIN for total RNA extracted from both axis and cotyledon of all varieties of (A) chickpea, (B) pea and (C) lentil stored at 2°C and 20°C for 20 years. The solid line represents a 1:1 ratio if RIN from axis and cotyledon tissue from a variety are the same. The dotted line represents the linear relationship between embryonic axis and cotyledon tissue within each crop.
Relationship of RIN with germination parameters
The rate of RIN decline in stored seeds is affected by the temperature and moisture of the storage environment (Tables 4 and 5). RIN decline rate also varies at the species, cultivar and tissue levels. Is there a relationship between variation in RIN decline with variation in the rate of viability loss among species and cultivars? To address this question, we expressed the rate of viability decline as final germination percentage, P50 and P50−1, considering that all samples had similar, high, initial germination (98–100% germination), and were stored under similar conditions (20°C and 30–40% RH) for the same time (20 years). We expressed RIN decline as average RIN of either the axis or the cotyledon (presuming RIN values of freshly harvested seeds were similar and >8) or as the difference in RIN between 2016 and 2023 measurements when both years were tested (RIN2016–2023). When all available samples from all species were considered in the models (N ≥ 20), RIN decline was significantly associated (P < 0.01) with the three measures of viability loss in eight of the nine models (Table 6). The strongest relationships (R 2 = 0.470, P < 0.0001, n = 22) used RIN values from RNA extracted from cotyledons and viability decline expressed as either germination in 2023 or as P50−1 (Fig. 3).
Table 6. Correlations within and across species between viability parameters (germination percentage, P50 or P50−1), and RIN from embryonic axis tissue, cotyledon tissue and RIN difference between 2016 and 2023 (Table 5)
Shaded rows indicate a significant correlation. The top portion of the table included analyses conducted with data from seeds stored at 20°C. The lower portion of the table includes germination percentage and RIN from 2023 for seeds stored at 20 and 2°C.
Fig. 3. Correlation of RIN with germination parameters of (A) ageing rate (P50−1) and (B) final germination percentage for chickpea, pea and lentil in embryonic axis tissue and cotyledon tissue. (A) Data from seeds stored at 20°C and (B) data from seeds stored at both the 2 and 20°C experiments. Open symbols represent RIN values collected from embryonic axis tissue; closed symbols represent RIN values collected from cotyledons. Each data point represents an individual variety within crop included in the study (Table 1). P50 was extrapolated using the Avrami function for germination data for each variety over the 20-year experiment and P50−1 was used to quantify the rate at which viability was lost under experimental conditions for each crop variety. Final germination is from 2023, except for samples that were depleted (Table 3), in which case 2016 final germination and RIN are depicted. Regression lines represent models including all species, solid lines represent regressions for cotyledon tissue and dashed lines represent regressions for embryonic axis tissue.
Regression analyses were weaker when restricted to within-species analyses. There were no significant relationships uncovered for chickpea seeds, which is to be expected because the four cultivars tested showed little sign of deterioration during 20 years of storage under the study conditions (Table 6). For pea, linear regressions between RIN from cotyledons and germination percentage in 2023 and P50−1 were significant (R 2 = 0.44, P = 0.036, n = 9 and (R 2 = 0.39, P = 0.055, n = 9). For lentil, linear regressions between RIN from cotyledons and P50−1 or RIN2016–2023 and germination percentage in 2023 were significant (R 2 > 0.67, P < 0.017, n = 9). Including germination and RIN data from the 2°C experiment further strengthened relationships (Table 6) of pea and lentil, but the relationship between RIN and viability remained not significantly different in chickpea, presumably because most of the seeds showed little sign of deterioration.
Discussion
We seek markers of seed ageing during storage that are expressed before seeds die but portend the timing of seed death. This paper focuses on RIN as a candidate marker because RNA accumulated during seed development tends to fragment during dry storage and can be detected before seed mortality (Fleming et al., Reference Fleming, Richards and Walters2017, Reference Fleming, Hill and Walters2019; Tetreault et al., Reference Tetreault, Fleming, Hill, Dorr, Yeater, Richards and Walters2022). The study requires materials at various stages of degradation from known (i.e. storage environment) and unknown (i.e. initial seed quality) mechanisms. An experiment begun in 2003 provides an opportunity to explore RIN in seeds from different cultivars of chickpea, pea and lentil that were stored under controlled moisture levels at 2 and 20°C and appear to be expressing variation in longevity. We demonstrate a decline in RIN that corresponds with seed ageing.
Intuitively, we know that a seed can germinate until it cannot. The binary nature of viability creates some philosophical problems when monitoring germination potential of a population over time. Some models predict a progressive, yet oblique, decline in germination initially, based on the assumption that seed deaths are normally distributed in time, with a caution of the limitation of applying models to other seed lots with variations in storage conditions (Ellis and Roberts, Reference Ellis and Roberts1980; Ellis, Reference Ellis1988). Alternatively, viability per se may be an inappropriate indicator of ageing if ageing results from an accumulation of minor damage that does not initially affect a seed's ability to germinate. In that scenario, measuring accumulated damage, as well as how much is required to induce a major effect, can lead to a more mechanistic understanding of the ageing process as well as better tools to measure it. In this case, models such as the Avrami equation (Avrami, Reference Avrami1941; Niedzielski et al., Reference Niedzielski, Walters, Luczak, Hill, Wheeler and Puchalski2009) which consider cooperativity and thresholds leading up to a transition can be helpful (Fig. 1). With either scenario, a huge issue for lab testing is how a small number of seeds in a test, as well as inconsistency among tests, will limit detection of any subtle decline in germination percentage (Tetreault et al., Reference Tetreault, Fleming, Hill, Dorr, Yeater, Richards and Walters2022). For example, at least 68 seeds are needed for statistical confidence that a decline from 98 to 90% germination is real (Tetreault et al., Reference Tetreault, Fleming, Hill, Dorr, Yeater, Richards and Walters2022). Data from this study show the common observation that germination results sometimes fluctuate among tests separated by storage time, making it difficult to interpret a high and low germination result in the vicinity of the mortality threshold (Table 3).
Parameters that quantify viability and RIN decline during storage
Uncertainty about the reliability of various parameters to quantify longevity prompted us to use several options to explore significant relationships (Hay et al., Reference Hay, Valdez, Lee and Sta. Cruz2019). The most familiar method uses germination percentage after a certain time. The time of 20 years achieved samples that varied in germination percentage. A shorter or longer storage time could have yielded all populations germinating above 90% or below 20%, respectively, which would not be informative. However, longevity is most directly assessed as a time parameter, and this can be obtained by fitting the monitor data time series to a sigmoidal model, as we have done using the Avrami equation (Fig. 1). Our longevity calculations consider the time for germination to decrease to 50%, P50. The use of all the 20-year germination data (Redden and Partington, Reference Redden and Partington2019; Table 3) evens out variation implicit in individual tests and provides a more comprehensive view than a single germination test of the storage behaviour of the sample. That said, uncertainty in P50 increases in samples that show minor to no degradation; there is a poor fit to the sigmoidal curve and P50 values tend towards infinity when germination percentage equals initial values. This problem is mitigated by expressing the reciprocal of P50 (P50−1), which transforms a very large P50 (i.e. long-lived seed) to an ageing rate that approaches 0 (i.e. slow ageing seed). In this study of 20-year-old seeds from different cultivars of chickpea, pea and lentil, 11 of the 30 cultivars stored at 20°C exhibited P50 > 20 years, meaning that about one-third of the samples showed minor to no viability decline. Presumably, P50 at 2°C ≫ P50 at 20°C, but we were unable to estimate P50 at 2°C given just two germination tests 20 years apart that were not significantly different (Hay et al., Reference Hay, Valdez, Lee and Sta. Cruz2019, Reference Hay, Davies, Dickie, Merritt and Wolkis2022). After 20 years of storage at 20°C, germination percentages ranged from 0 to 100%; P50 ranged from 11 to 68 years; and P50−1 ranged from 0.090 to 0.015 yr−1 (or −4.5 to −0.7% germ yr−1).
We also provide three options for how the RIN (a quantitation of the amount of fragmentation of RNA molecules) parameter is expressed. Large-seeded legumes can be easily separated into nutritive tissues (i.e. cotyledons) and tissues that grow and develop into the seedling (i.e. embryonic axis). High quality RNA can be extracted from both tissue types in fresh seeds, but RNA yield per mg tissue is >10-fold higher from axes compared with cotyledons (unpublished and Fleming et al., Reference Fleming, Richards and Walters2017). We also had the opportunity to compare RIN values from seeds that were removed from storage 7 years apart (2016 and 2023, RIN2016–2023) to estimate change in RIN between monitor tests. On average, RIN values were significantly lower (P = 0.0186) in samples removed from storage in 2023 compared with 2016 (5.5 ± 1.5 (n = 18) and 5.9 ± 1.2 (n = 22), respectively). The average rate of change of RIN was about 0.07 RIN⋅yr−1, which leads to a predicted starting value of ≈6.9 (6.9 = 5.5 RIN + 0.07 RIN⋅yr−1 * 20 yr) for these samples. This prediction is lower than average RIN values measured in 2023 for seeds stored at 2°C (RIN ≈ 8), which suggests that RIN declined in 20°C-stored seeds faster than we estimated.
Correlations between viability and RIN decline during storage
We used regression analysis to test the hypothesis that RNA fragmentation occurs at comparable rates as the decline in germination percentage, P50 or P50−1, making it a good marker of seed ageing. The hypothesis is based on the supposition that ageing is caused by incremental damage to cellular constituents that eventually lead to seed death. Moreover, the kinetics of damaging reactions are controlled by factors that affect molecular mobility in the cytoplasm (namely moisture and temperature), as well as substrate levels for reactants (e.g., reactive oxygen molecules) and protectants or ‘decoys’ for damaging oxidative reactions (Roberts, Reference Roberts1973; Ballesteros and Walters, Reference Ballesteros and Walters2019; De Vitis et al., Reference De Vitis, Hay, Dickie, Trivedi, Choi and Fiegener2020; Solberg et al., Reference Solberg, Yndgaard, Andreasen, Von Bothmer, Loskutov and Asdal2020). We evaluated the relationship between RIN and viability through regression analysis across all three species and within species. RIN × germination percentage data for the 2°C storage experiment was added to the correlation analysis to test the overall hypothesis of RIN decline with storage temperature and moisture data. When the correlation model combined all species and all treatments, the significance of the relationship was high (P ≤ 0.0136, R 2 ≥ 0.21) for all cross comparisons except RIN2016–2023 versus P50 (Table 6). Relationships using P50 and RIN2016–2023 were weakest, probably because P50 was overestimated in samples with high viability in 2023 and RIN2016–2023 included just a subset of all the samples. When correlation models using these factors were removed, P ≤ 0.0002 and R 2 ranged between 0.39 and 0.88 (Table 6; Fig. 3).
Relationships were expectedly weaker when individual species were considered because N was smaller and within-species variation in longevity was lower than variation across all species. In fact, a significant correlation was not gleaned for chickpea and this is likely because just two of the samples included in RIN assays showed significant indication of viability decline (i.e. cv. Paider-91 at 20°C and Howzat at 2°C, high moisture). For pea and lentil, correlations were significant when RIN from cotyledons were regressed against P50−1; including data from the 2°C slightly aged samples strengthened the relationships (P < 0.0001, R 2 > 0.63) (Table 6; Fig. 3).
Significant relationships between viability loss and RNA fragmentation suggest these two manifestations of seed ageing interact. However, the initial slow change of viability (Fig. 1) is hard to detect, whereas the presumed linear decline in RIN (not examined here, but see Fleming et al., Reference Fleming, Hill and Walters2019; Tetreault et al., Reference Tetreault, Fleming, Hill, Dorr, Yeater, Richards and Walters2022) means that RNA fragmentation can be detected before significant mortality. For example, a monitoring interval of 10 years precedes P50 for all samples studied but in contrast would show a very detectable average RIN decline of about 0.7.
Variation in seed longevity among species, among cultivars and between tissue types
Temperature and moisture effects on seed longevity are increasingly understood, especially in the context of mobility within vitrifying cytoplasm (Walters, Reference Walters1998; Walters et al., Reference Walters, Ballesteros and Vertucci2010; Ballesteros and Walters, Reference Ballesteros and Walters2019). This study provides a unique dataset examining an elusive factor of initial seed quality that is most likely related to the chemical composition and organization of cytoplasm in dried cells. Seeds acquire longevity during the embryogenic programme (Hay et al., Reference Hay, Probert and Smith1997; Righetti et al., Reference Righetti, Vu, Pelletier, Vu, Glaab, Lalanne, Pasha, Patel, Provart, Verdier, Leprince and Buitink2015; Whitehouse et al., Reference Whitehouse, Hay and Ellis2015; Pereira Lima et al., Reference Pereira Lima, Buitink, Lalanne, Rossi, Pelletier, da Silva and Leprince2017; Zinsmeister et al., Reference Zinsmeister, Leprince and Buitink2020) and this capacity is often expressed as a species characteristic (Priestley et al., Reference Priestley, Cullinan and Wolfe1985; Roos and Davidson, Reference Roos and Davidson1992; Walters et al., Reference Walters, Wheeler and Grotenhuis2005; Ellis and Hong, Reference Ellis and Hong2006; Probert et al., Reference Probert, Daws and Hay2009; Merritt et al., Reference Merritt, Martyn, Ainsley, Young, Seed, Thorpe, Hay, Commander, Shackelford, Offord, Dixon and Probert2014; Solberg et al., Reference Solberg, Yndgaard, Andreasen, Von Bothmer, Loskutov and Asdal2020). In the few studies that compare seed longevity among species, chickpea, pea and lentil are all considered to produce long-lived seeds (Priestley et al., Reference Priestley, Cullinan and Wolfe1985; Walters et al., Reference Walters, Wheeler and Grotenhuis2005; Solberg et al., Reference Solberg, Yndgaard, Andreasen, Von Bothmer, Loskutov and Asdal2020). Chickpea seeds stored at low water content (~5%) in hermetically closed packets stored at −18°C showed relatively stable storage after 20–24years with a less than 5% viability decline compared with initial viability (Desheva, Reference Desheva2016). Here, the average P50 of chickpea seeds stored at 20°C was 24.9 years and it was predicted to be 28 years from the viability equations using comparable environmental conditions (SER, 2023). Pea seeds are also identified as long-lived with a P50 ~ 100 years under dry room conditions or genebanks (Roos and Davidson, Reference Roos and Davidson1992; Solberg et al., Reference Solberg, Yndgaard, Andreasen, Von Bothmer, Loskutov and Asdal2020). Here, the average P50 of pea seeds stored at 20°C was 19.5 years and the viability equations predicted a similar longevity as chickpea at 28 years (SER, 2023). Lentil seeds stored at NLGRP had an average P50 of 365 years (Walters et al., Reference Walters, Wheeler and Grotenhuis2005); however, here (and in Priestley et al., Reference Priestley, Cullinan and Wolfe1985), they were less stable than the other two species with an average P50 of 17.1 years and a predicted P50 of 19 years based on the viability equations and new species coefficients (Whitehouse and Norton, Reference Whitehouse and Norton2022).
Differences in average longevity among these species are not surprising. As we have described, prediction of longevity of long-lived material requires extrapolation and associated uncertainty. Variation among labs is also likely due to the selection of cultivars used in the study. There has been increasing awareness of within-species variation in seed longevity that is due to genetic background and growth conditions (Clerkx et al., Reference Clerkx, El-Lithy, Vierling, Ruys, Blankestijn-De Vries, Groot, Vreugdenhil and Koornneef2004; Nagel et al., Reference Nagel, Rosenhauer, Willner, Snowdon, Friedt and Börner2011, Reference Nagel, Kranner, Neumann, Rolletschek, Seal, Colville, Fernández-Marín and Börner2015; Lee et al., Reference Lee, Velasco-Punzalan, Pacleb, Valdez, Kretzschmar, McNally, Ismail, Cruz, Sackville Hamilton and Hay2019), rather than minor differences in germination capacity as modelled in the viability equations. In this study, we observed considerable variation in longevity among legume grain cultivars that were grown, harvested and stored similarly. While the underlying basis for differing longevities of seeds within a species cannot be deciphered from this study, the correlation between RIN and eventual viability loss provides a promising approach to phenotyping the seed longevity trait without having to wait for seeds to die (Hay et al., Reference Hay, Valdez, Lee and Sta. Cruz2019).
Another novel aspect of this study is the report that the decrease in RIN appears more severe in embryonic axis tissue compared with cotyledons (Fig. 2). It was also surprising that the relationship of RIN with viability parameters (Table 6) was stronger for RNA extracted from cotyledons compared with the embryonic axis. This finding is consistent with a previous report for pea (Fleming et al., Reference Fleming, Hill and Walters2019), but differs from observations made using soybean seeds that showed no major differences between RIN measurements of RNA extracted from either cotyledons or embryonic axes (Fleming et al., Reference Fleming, Richards and Walters2017; Tetreault et al., Reference Tetreault, Fleming, Hill, Dorr, Yeater, Richards and Walters2022). We speculate that this combination of observations, as well as the extremely high yield of RNA from axes, may suggest that embryonic axes accumulate a superfluous supply of RNA and that there may be some variation in how RNA is protected within cells of the embryonic axes. These speculations can lead to inferences about RNA stability in cotyledons versus embryonic axes, the mechanism(s) of protection and the role of extraneous but more unprotected RNA molecules in fresh, mature seeds.
Consumption of seed during monitor testing
Hundreds of seeds were needed to conclusively show a decrease in germination with storage time (ISTA, 2003; Redden and Partington, Reference Redden and Partington2019; Tetreault et al., Reference Tetreault, Fleming, Hill, Dorr, Yeater, Richards and Walters2022) which was not really clear until seed germination was <[85–80%] in years [12–16] depending on species. In contrast, just 4–12 seeds provided an informative estimate of RIN and deterioration of seed health.
Conclusion
The viability of lentil, chickpea and pea seeds stored at 20°C began to decline after 16–20 years. RNA integrity also declined during this time, and RIN assessments reliably distinguished grain legumes that died sooner versus later. Separation was largely based on species, but also significantly correlated with aging rates of seeds from different genetic lines. Assaying RIN earlier in the experiment would likely have revealed variations of ageing rates among species and lines and might help to predict which seed lots would succumb first. RIN appears to be a promising marker for seed ageing that can aid genebanks in managing the difficult task of predicting and detecting imminent seed deaths. We conclude that RNA integrity (RIN) is a feasible approach to assay the progress of seed ageing. RIN assays can be interspersed with germination assays to aid interpretation of germination assays and adjust monitoring or regeneration schedules to ensure genebanks maintain high quality germplasm. Future work correlating the rate of RIN decline with deterioration time courses may lead to more reliable prediction of the onset of rapid mortality of stored seeds.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0960258524000072.
Acknowledgements
Thanks to Debra Partington for helping with the initial set-up of this 20-year experiment. Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the U.S. Department of Agriculture (USDA). USDA is an equal opportunity provider and employer.
Conflicts of interest
The authors declare none.
Open access
