Volatile fingerprints of seeds of four species indicate the involvement of alcoholic fermentation, lipid peroxidation, and Maillard reactions in seed deterioration during ageing and desiccation stress

Abstract

The volatile compounds released by orthodox (desiccation-tolerant) seeds during ageing can be analysed using gas chromatography–mass spectrometry (GC-MS). Comparison of three legume species (Pisum sativum, Lathyrus pratensis, and Cytisus scoparius) during artificial ageing at 60% relative humidity and 50 °C revealed variation in the seed volatile fingerprint between species, although in all species the overall volatile concentration increased with storage period, and changes could be detected prior to the onset of viability loss. The volatile compounds are proposed to derive from three main sources: alcoholic fermentation, lipid peroxidation, and Maillard reactions. Lipid peroxidation was confirmed in P. sativum seeds through analysis of malondialdehyde and 4-hydroxynonenal. Volatile production by ageing orthodox seeds was compared with that of recalcitrant (desiccation-sensitive) seeds of Quercus robur during desiccation. Many of the volatiles were common to both ageing orthodox seeds and desiccating recalcitrant seeds, with alcoholic fermentation forming the major source of volatiles. Finally, comparison was made between two methods of analysis; the first used a Tenax adsorbent to trap volatiles, whilst the second used solid phase microextraction to extract volatiles from the headspace of vials containing powdered seeds. Solid phase microextraction was found to be more sensitive, detecting a far greater number of compounds. Seed volatile analysis provides a non-invasive means of characterizing the processes involved in seed deterioration, and potentially identifying volatile marker compounds for the diagnosis of seed viability loss.

Introduction

Plant diversity is being lost at an unprecedented rate, and conservation of plant species is now a global priority. Ex situ conservation becoming the only viable option for preserving plant species that otherwise face the threat of extinction due to habitat loss, and it is estimated that 90% of the 6 million accessions of plant genetic resources held globally reside in seed banks (Li and Pritchard, 2009). The challenge with storing large and diverse seed collections is that ageing is inevitable, even when seeds are stored at low humidity and temperature, and conventional germination testing to assess seed viability can be time-consuming and destroys germplasm. There is also great variation in the longevity of seeds between species, which makes the prediction of storage lifespan difficult (Pritchard and Dickie, 2003). A non-invasive technique for monitoring seed viability would save precious seed collections, and may also serve as an early warning of impending viability loss to allow intervention, for example regeneration of seed collections, to take place. One potential non-invasive technique is to use gas chromatography–mass spectrometry (GC-MS) to analyse the air within seed storage containers.

A link between seed deterioration and the release of volatile compounds has been reported in several studies. Hailstones and Smith (1989) observed a significant correlation between seed germination and vigour and the relative abundance of volatile aldehydes released by soybean seeds that had been heated to 130 °C for 2h. The volatile aldehydes were proposed to derive from the thermal degradation of lipid hydroperoxides. Volatile aldehydes were also released by soybean seeds following deterioration due to weathering, but were not evolved from dry, non-germinating, and dead seeds, so it was concluded that active metabolism was required for their production (Tyagi, 1992). However, other reports indicate that volatile production does occur from dry seeds. Zhang et al. (1993) analysed the volatiles released by the dry seeds of 47 species from several different families during storage at 23 °C. A total of 24 volatile compounds were detected, and the number of volatile compounds and their abundance increased with storage time. Ethanol, acetaldehyde, methanol, and acetone were detected in most species and, since ethanol and acetaldehyde are intermediates of glycolysis, it was postulated that metabolic reactions were taking place, albeit very slowly, even in dry seeds. No volatiles were detected when seeds were stored at –10 °C, but the release of volatiles could be stimulated by heating at 80 °C for 20min. Zhang et al. (1993) proposed that seeds produced volatile compounds during low temperature storage but that the release of volatiles was suppressed. However, heating to 80 °C will also accelerate seed deterioration, and hence increase volatile emissions.

Volatile production by dry seeds is associated with degradative processes such as lipid peroxidation during ageing (Zhang et al., 1995a ). However, the mechanisms of seed ageing during storage at low moisture content (MC) and low temperature have yet to be elucidated. Under such storage conditions, metabolic activity is minimal due to the low molecular mobility of cell cytoplasm. One mechanism of seed ageing is via Maillard reactions, which occur following a non-enzymatic attack on amino groups of proteins and nucleic acid–protein complexes by reducing sugars or aldehydes (Murthy and Sun, 2000). This forms advanced glycosylation end-products, which are found in both naturally and artificially aged seed tissues, and increase during storage at all MCs and temperature conditions. The rate of Maillard reactions increases with temperature and relative humidity, and the accumulation of Maillard reaction products is associated with loss of seed viability (Wettlaufer and Leopold, 1991). The modification of macromolecules through Maillard reactions affects protein activity, in particular some antioxidant enzymes such as glutathione reductase, ascorbate peroxidase, and catalase are sensitive to Maillard reactions, which causes a decline in the antioxidant capacity, and an inability to limit oxidative damage during germination, resulting in impaired seed vigour and loss of viability (Murthy et al., 2003).

Two broad seed storage types have been distinguished based on their tolerance to desiccation: orthodox (desiccation-tolerant) and recalcitrant (desiccation-sensitive) seeds. Generally, orthodox seeds undergo dehydration during development on the mother plant and are desiccation tolerant when shed at maturity. The low MC of mature orthodox seeds prevents the formation of ice crystals on cooling; thus orthodox seeds are highly tolerant of subzero temperatures. In contrast, recalcitrant seeds (Farnsworth, 2000; Berjak and Pammenter, 2008) are desiccation sensitive and intolerant of freezing temperatures as whole seeds. They are shed at high MC and remain metabolically active until germination. The sensitivity of recalcitrant seeds to desiccation is thought to be due to metabolic imbalance as the water content falls, which results in formation of reactive oxygen species (ROS). Desiccation-tolerant seeds undergo a concerted down-regulation of metabolism during maturation drying, and therefore limit ROS accumulation (Leprince et al., 2000). Dehydration of desiccation-sensitive tissues of pea and cucumber radicles was associated with the production of acetaldehyde, indicative of a switch to fermentation, possibly triggered by a disruption of flux through glycolysis, the tricarboxylic acid cycle, or oxidative phosphorylation (Leprince et al., 2000). Volatile production during deterioration of recalcitrant seeds is not widely reported in the literature, although a study of four recalcitrant species (Ligustrum japonicum, Quercus serrata, Quercus myrsinaefolia, and Camellia japonica) found that seeds produced methanol, ethanol, and acetaldehyde during storage, and that the application of exogenous acetaldehyde accelerated deterioration (Akimoto et al., 2004).

There are several approaches for analysing volatile compounds using GC-MS. Static headspace analysis involves sealing the sample into a container and measuring the emitted volatiles. The headspace can be sampled directly using a gas-tight syringe (Zhang et al., 1995a ; Mira et al., 2010) or the volatiles can be concentrated using an adsorbent trap, for example Tenax, a polymer resin (Zhang et al., 1993). Solid phase microextraction (SPME) uses an adsorbent fibre within a syringe. The fibre is exposed to the headspace within a vial, and the trapped volatile compounds are then transferred to a GC-MS by thermal desorption (Tholl et al., 2006; Mira et al., 2010). The above sampling techniques have all been used previously in studies of seed volatiles, but there are few reports of comparison between the techniques.

The aim of this study was to characterize volatile production during artificial ageing of orthodox seeds of three species (Pisum sativum, Lathyrus pratensis, and Cytisus scoparius) belonging to the Leguminosae family. Pisum sativum is a crop legume, whilst the other two are wild species. Comparison of closely related species increases the likelihood of identifying common volatile compounds and therefore potential candidates for viability markers that can then be assessed with a greater range of species. Artificial ageing was used as a convenient means of mimicking natural ageing, although the exact mechanisms and kinetics of ageing may differ. The volatile signature of aged seeds was compared with that of a recalcitrant seed (Quercus robur) during desiccation to see whether the cause of seed deterioration influenced the volatile compounds produced. In addition, analyses of fatty acid composition and lipid peroxidation products were performed to determine whether volatile production correlated with lipid peroxidation. Comparison was made between two methods of analysis; the primary data were obtained using Tenax, which was exposed to the seeds throughout the ageing/desiccation treatments. Seeds were subsequently re-analysed using SPME. Due to the constraints on sample size using SPME [the maximum vial size accepted by the CombiPAL autosampler (CTC Analytics AG, Switzerland) was 20ml, with an opening diameter of 12 mm], whole Q. robur seeds could not be analysed, and therefore subsamples of all species were powdered prior to SPME analysis.

Materials and methods

Seed material and controlled deterioration

Seeds of P. sativum L. Alaska Early (Abundant Life Seed Foundation, Port Townsend, USA), L. pratensis L. (B&T World Seeds, France), and C. scoparius L. were commercially obtained. Seeds of Q. robur were collected from the grounds of the Royal Botanic Gardens, Kew, Wakehurst Place. The seed coats of C. scoparius and L. pratensis were chipped to remove physical dormancy. All orthodox seeds were equilibrated over lithium chloride at 60% relative humidity (RH) prior to ageing at 50 °C in sealed 500ml (P. sativum) or 250ml (C. scoparius and L. pratensis) glass bottles (ensuring that the volume of seeds:air was ~1:20) containing 500mg of Tenax (Supelco, Gillingham, UK) in 10ml glass headspace vials. The ageing period was adapted for each species: P. sativum (0, 8, 12, 15, 25, and 55 d); L. pratensis (0, 35, and 70 d); and C. scoparius (0, 21, 49, and 63 d). Quercus robur acorns were desiccated over 1.2kg of silica gel at 20 °C for 0, 3, 6, and 13 d in sealed 3 litre glass jars containing 500mg of Tenax in 10ml headspace vials. Five replicate samples for each ageing/desiccation time point were prepared in separate bottles for P. sativum (280 seeds per replicate) and Q. robur (160 seeds per replicate), whilst four replicates were used for L. pratensis (195 seeds per replicate) and C. scoparius (255 seeds per replicate). In addition, blank samples of Tenax were stored in glass bottles at 50 °C for 7 d. Seed viability and MC were assessed as described by Kranner et al. (2006).

GC-MS analysis of seed volatiles trapped using Tenax adsorbent

Following seed ageing or desiccation, the headspace vials containing the Tenax were removed from the glass bottles, sealed, and stored at room temperature prior to analysis. A 50mg aliquot of the Tenax was subsampled from each vial, transferred to a 10ml headspace vial, and 0.4 µg of d₁₀-diethyl ether (internal standard; Cambridge Isotope Laboratories, Cambridge, UK) in 1% Tween (Sigma-Aldrich, Gillingham, UK) was added. The vials were incubated at 75 °C for 50min prior to analysis by headspace GC-MS (Voyager system, Thermoquest, UK) using an Rt-QPLOT column (30 m length, 0.32mm internal diameter, 10 µm film thickness; Restek, UK) with a temperature program (2min at 40 °C, 5 °C min^–1 to 250 °C, 5min hold). A splitless (1min) injection of 1000 µl was performed with a heated syringe (90 °C). The injection port and transfer line temperatures were 200°C and 250°C, respectively. The GC-MS was operated in full scan mode (30 - 300 amu). A series of calibration standards (0–200 µg g^–1 Tenax) of ethanol, 2-propanol, pentane, 1-propanol, 2-methyl pentane, 3-methyl pentane, hexane, butanol, and heptane were prepared in 1% Tween and spiked onto 50mg of Tenax. For some low abundance volatile compounds, 100mg of Tenax was subsampled and the GC-MS was operated in selected ion monitoring (SIM) mode: 2-methylpropanal (m/z 41, 43, and 72), 2,3-butanedione (m/z 43 and 86), methyl acetate (m/z 43, 59, and 74), 2-methylbutanal (m/z 57, 58, and 86), and d₁₀-diethyl ether (m/z 66 and 84). Calibration standards for these compounds were spiked onto 100mg of Tenax (0–20 µg g^–1 Tenax).

Analysis of aged Pisum sativum, Cytisus scoparius, and Lathyrus pratensis seeds and desiccated Quercus robur acorns using solid phase microextraction and GC-MS

Seeds were freeze-dried and ground to a powder using a laboratory dismembrator (Retsch, UK). A 1g aliquot was placed into 20ml headspace vials. Headspace sampling was performed using automated SPME with a 75 µm Carboxen-PDMS fibre (Supelco, Gillingham, UK) and an extraction time of 30min at 30 °C followed by 5min desorption in the GC injector port at 240 °C. The volatiles were separated using GC (Thermo Finnigan Trace GC Ultra) on a FAMEWAX column (30 m length, 0.25mm internal diameter, 0.25 µm film thickness; Restek, UK) running a temperature program (3min hold at 35 °C, 3 °C min^–1 to 60 °C, 10 °C min^–1 to 220 °C, and 1min hold; helium carrier gas at constant flow rate of 1ml min^–1). The volatiles were detected using MS (Thermo Finnigan Trace DSQ; ionization energy 70eV, scan frequency range m/z 10–350 per 0.5 s), and identified from the NIST mass spectral database. Identities were confirmed with analytical standards (Sigma-Aldrich, Gillingham, UK): acetaldehyde, acetic acid, acetone, acetonitrile, benzaldehyde, butanal, 2-butanol, 2-butanone, 3-carene, ethanol, ethylbenzene, heptanal, hexanal, n-hexylformate, d-limonene, methanol, 2-methylbutanal, 3-methylbutanal, 2-methyl-1-propanol, nonanal, 1-octen-3-ol, pentanal, 1-pentanol, 1-penten-3-ol, propanal, 1-propanol, 2-propanol, 2-propenenitrile, and xylene. To account for variation in the detector response to different compounds, response factors were calculated by dividing the peak area by the concentration of analytical standards over a range of concentrations. The response factors were then used to calculate the amount of each volatile compound present in the headspace of the samples.

Analysis of seed fatty acids using GC-MS

A 20–50mg aliquot of seed powder was mixed with 50mg l^–1 butylated hydroxytoluene (BHT) in 6.4ml of isopropanol and 10 µl of 10mg ml^–1 heptadecanoic acid (internal standard), and centrifuged at 2300rpm and 4 °C for 2min. The pellet was re-extracted with 4ml of 2-propanol containing 50mg ml^–1 BHT and 4ml of chloroform overnight at room temperature with constant shaking (150rpm). After centrifugation, the supernatants were combined and evaporated to dryness at 45 °C under a nitrogen stream. The residue was dissolved in 2ml of chloroform:methanol [2:1 (v/v)] and 0.5ml of 0.88 % (w/v) potassium chloride, shaken, and the upper phase was removed. The lower phase was washed with 0.7ml of methanol:0.88 % (w/v) potassium chloride [1:1 (v/v)] and dried under nitrogen at 45 °C. The residue was re-suspended in 1ml of toluene, and fatty acids were derivatized with 2ml of 1% sulphuric acid in methanol overnight at 50 °C. Excess derivatizing agent was removed by washing with 5ml of hexane and 5ml of 5% (w/v) sodium chloride. The lower phase was re-washed with 5ml of hexane, and the combined hexane phases were evaporated under nitrogen at 45 °C. The residue was dissolved in 1ml of hexane and transferred to a 2ml autosampler vial for GC-MS analysis. A split injection (100:1 split ratio) of 1 µl was injected onto the GC (Thermo Finnigan Trace GC Ultra), and fatty acid methyl esters were separated using a FAMEWAX column (30 m length, 0.25mm internal diameter, 0.25mm df; Restek, UK) running a temperature program (initial temperature 70 °C, 20 °C min^–1 to 195 °C, 5 °C min^–1 to 240 °C, 10min hold at 240 °C; helium carrier gas at a constant flow rate of 1ml min^–1). The compounds were detected using MS (Thermo Finnigan Trace DSQ; ionization energy 70eV, scan frequency range m/z 10–500 per 0.3 s), and identified through comparison with the NIST mass spectral database and analytical standards (F.A.M.E. Mix C4–C24, Supelco). Quantification of fatty acid methyl esters was performed using standard curves of quantitative standard mixtures (F.A.M.E. Mix GLC-10, -30, and -50, Supelco).

Determination of malondialdehyde and 4-hydroxynonenal in Pisum sativum seeds

The levels of malondialdehyde and 4-hydroxynonenal were determined according to a method described by Muckenschnabel et al. (2001) using high-performance liquid chromatography with atmospheric pressure chemical ionization mass spectrometric detection (LC-APcI-MS). Separation was performed on a Primesphere C18 HS column (5 µm particle size, 250mm length, 3.2mm diameter; Phenomenex) and malondialdehyde and 4-hydroxynonenal derivatives were detected using a VG Platform mass spectrometer (Micromass, Manchester, UK) operating in negative ion APcI mode. Selected ion monitoring of ions with m/z of 274, 234, 204, and 117 for malondialdehyde; 417 and 335 for 4-hydroxynonenal; and 303, 256, and 139 for 4-fluorobenzaldehyde (internal standard) was used.

Results

Effect of controlled deterioration and desiccation treatment on seed viability

Controlled deterioration resulted in a loss of seed viability, which was assessed through germination testing. In P. sativum (Fig. 1A), ageing at 50 °C for up to 15 d had no effect on total germination, with 100% of seeds aged for 8, 12, and 15 d germinating, compared with initial germination of 96% in non-aged seeds, although aged seeds were slower to start germinating. After 25 d of ageing, total germination was reduced to 85%, and viability was almost completely lost after 55 d of ageing. Ageing of C. scoparius seeds (Fig. 1B) reduced total germination from 88% in non-aged seed to 58, 63, and 50% in seeds aged for 21, 49, and 63 d, respectively. Again germination was slightly slower in the aged seeds. Lathyrus pratensis viability (Fig. 1C) was 93% in non-aged seed, and this fell to 40% and 8% in seeds aged for 35 d and 70 d, respectively. Non-aged seeds completed germination within 8 d, compared with 10 d and 15 d in seeds aged for 35 d and 70 d. The initial viability of Q. robur acorns (Fig. 1D) was 50%, which although low is not unusual for field-collected seed lots. The MC was 46%. Desiccation for 3 d reduced the MC to 45% and slowed the germination rate, although total germination increased to 62%. Desiccation for 6 d to 43% MC further slowed germination, and viability was reduced to 34%. After 13 d of desiccation, the MC was 37% and total germination had fallen to 3%.

Fig. 1.

Ageing of Pisum sativum, Lathyrus pratensis, and Cytisus scoparius seeds, and desiccation of Quercus robur seeds caused loss of viability, shown by a decline in total germination and a reduction in seed vigour, indicated by an increase in the amount of time taken to complete germination. The graphs show the total germination of P. sativum (A) seeds following ageing for 0 (white circles), 8 (grey circles), 12 (black circles), 15 (white triangles), 25 (grey triangles), and 55 d (black triangles); C. scoparius (B) seeds aged for 0 (white circles), 21 (white triangles), 49 (grey triangles). and 63 d (black triangles); L. pratensis (C) seeds aged for 0 (white circles), 35 (grey triangles), and 70 d (black triangles); and Q. robur (D) seeds desiccated for 0 (white circles), 3 (white triangles), 6 (grey triangles), and 13 d (black triangles). Each data point represents the mean ±SE of four (C. scoparius and L. pratensis) or five (P. sativum and Q. robur) replicates. Letters indicate statistically significant differences (P < 0.05) in total germination between ageing/desiccation treatments.

Volatile compounds released during deterioration of seeds

Several volatile compounds were trapped in the Tenax exposed to P. sativum seeds (Table 1, Fig. 2A) during ageing. Ethanol, methyl acetate, 2-propanol, 2,3-butanedione, and 2-methylpropanal were found only in Tenax from aged seeds, and the concentrations increased with seed ageing time. Methyl acetate was the most abundant volatile, followed by ethanol, which appeared in seeds aged for 8 d and then increased 5-fold in seeds aged for 55 d. Ethanol, 2-propanol, methyl acetate, and 2-methylpropanal were also produced by C. scoparius seeds (Table 1, Fig. 2B) during ageing. Other volatiles released by aged C. scoparius seeds were 2-methyl-2-propanol, hexene, and 2-methylbutanal, the latter being the most abundant. Despite the greatest loss of viability occurring between 0 and 3 weeks of ageing, volatile production showed the greatest increase between 3 and 6 weeks of ageing. The relationship between the ageing period and volatile production was generally weaker in C. scoparius than in P. sativum, and there were no significant relationships between viability and production of individual volatiles (Table 1). Four compounds evolved from ageing L. pratensis seeds (Fig. 2C): methyl acetate, which was the most abundant, followed by 2-methylbutanal, 2-methylpropanal, and 2,3-butanedione. However, none correlated with ageing or viability. The compounds that were common to all three species were methyl acetate and 2-methylpropanal. Volatile compounds were also released from acorns (Fig. 2D) during desiccation. Ethanol was the most abundant compound and was released only by acorns that had been desiccated. Production increased linearly at 3 d and 6 d of desiccation, and then increased sharply in seeds that had lost viability after 13 d. 2,3-Butanedione was also released by acorns, along with 2-pentanone and one unidentified compound. To investigate whether the thick seed coat of acorns was affecting the release of volatile compounds, analysis of acorns with the seed coats removed was also performed. The removal of seed coats resulted in a >2-fold increase in the level of volatiles released by acorns desiccated for 13 d.

Table 1.

The volatile compounds released by ageing Cytisus scoparius, Lathyrus pratensis, and Pisum sativum seeds, and desiccating Quercus robur seeds were detected using a Tenax adsorbent and GC-MS. Correlation coefficients (R ²) indicate a significant (P < 0.05) relationship between ageing/desiccation period or germination and individual volatile compounds. Non-significant relationships are indicated by NS.

		Ageing/desiccation period		Germination (%)
		R 2	P-value	R 2	P-value
P. sativum (ageing)	1-Propanol	NS		NS
	2-Methylpropanal	0.531	<0.0001	0.479	<0.0001
	2-Propanol	0.505	<0.0001	0.458	<0.0001
	2,3-Butanedione	0.850	<0.0001	0.617	<0.0001
	Ethanol	0.497	<0.0001	0.548	<0.0001
	Methyl acetate	0.757	<0.0001	0.814	<0.0001
C. scoparius (ageing)	2-Methylpropanal	0.392	<0.01	NS
	2-Methyl-2-propanol	NS		NS
	2-Methylbutanal	0.460	<0.005	NS
	2-Propanol	NS		NS
	Ethanol	NS		NS
	Hexene	0.387	<0.05	NS
	Methyl acetate	0.272	<0.05	NS
L. pratensis (ageing)	2-Methylpropanal	NS		NS
	2-Methylbutanal	NS		NS
	2,3-Butanedione	NS		NS
	Methyl acetate	NS		NS
Q. robur (desiccation)	2-Pentanone	0.620	<0.0005	0.495	<0.005
	2,3-Butanedione	0.525	<0.005	0.420	<0.01
	Ethanol	0.586	<0.001	0.495	<0.005

Fig. 2.

The amount of volatile compounds released by the seeds increased with the ageing/desiccation period. The stacked bars represent the overall concentration of volatiles and volatile composition (indicated by different shading patterns) released by Pisum sativum (A), Cytisus scoparius (B), and Lathyrus pratensis (C) seeds during artificial ageing, and Quercus robur acorns during desiccation (D). Volatile compounds were trapped using Tenax and analysed using GC-MS. The lines and open circles represent the mean total germination at each ageing/desiccation time point. Each bar and data point represents the mean ±SE of five replicates for P. sativum and Q. robur, and four for C. scoparius and L. pratensis. The stacked bars corresponding to ‘seed’ on the x-axis of the Q. robur graph (D) represent the volatiles released after removal of the seed coat from Q. robur seeds that had been desiccated for 13 d. (This figure is available in colour at JXB online.)

Analysis of the volatile compounds released by powdered, aged seeds using SPME and GC-MS revealed more compounds than the initial headspace analysis with Tenax (Fig. 3, Table 2; Supplementary Fig. S1 available at JXB online). The production of some alcohols from P. sativum seeds was related to seed ageing and/or viability. Methanol showed the strongest relationship, whilst ageing appeared not to be related to ethanol production (Table 2). The levels of alcohols increased with ageing (Fig. 3F), and showed a stronger and more significant relationship with the ageing period than with viability (Table 2). In contrast, the production of volatile aldehydes increased at the earlier ageing time points, and reached a peak at 12 d of ageing, after which the levels started to decline (Fig. 3J). Other compounds detected by SPME which showed a significant relationship with seed ageing were acetic acid, ethylbenzene, and xylene, which increased as ageing progressed (Table 2, Fig. 3N; Supplementary Fig. S1). These were also detected in the headspace of ageing C. scoparius seeds, although only ethylbenzene and xylene showed a relationship with seed ageing, and in this case the levels declined during ageing (Fig. 3P; Supplementary Fig. S1). Several alcohols were detected by SPME, including 2-methyl-2-propanol, 2-propanol, and ethanol (Fig. 3H), which were also found in the analyses using Tenax (Tables 1, 2). Compounds which showed a significant relationship with seed ageing included 2-butanol, 2-propanol, diisobutylene, 3-carene, and d-limonene (Supplementary Fig. S1). Volatile emissions from L. pratensis seeds included several alcohols, of which methanol and ethanol were positively correlated with seed ageing, and 1-penten-3-ol declined during ageing (Fig. 3G; Supplementary Fig. S1). In addition, release of 2-butanone and acetic acid was negatively and positively correlated, respectively, with both ageing and viability (Table 2, Fig. 3O; Supplementary Fig. S1). The only compound detected by both SPME and Tenax was 2-methylbutanal (Tables 1, 2). SPME analysis of Q. robur acorns showed that ethanol was released during desiccation (Fig. 3A; Supplementary Fig. S1) and showed a strong correlation with both desiccation period and viability, which agreed with the results obtained using Tenax (Tables 1, 2). Methanol, 2-butanol, 1-propanol, and 2-methyl-1-propanol also increased with desiccation and viability loss (Fig. 3E; Supplementary Fig. S1). The aldehydes detected showed a similar response to that observed in P. sativum, with an initial increase during desiccation followed by a decline in the rate of production (Fig. 3I).

Fig. 3.

The volatile compounds present in the headspace of desiccated Quercus robur acorns and aged Pisum sativum, Lathyrus pratensis, and Cytisus scoparius seeds were analysed using solid phase microextraction (SPME). The seed material had been freeze-dried, ground, and stored at –75 °C prior to analysis. The stacked columns represent the total volatile concentration and composition (indicated by different shading patterns) released by Q. robur acorns following desiccation (A, E, I, M), and P. sativum (B, F, J, N), L. pratensis (C, G, K, O), and C. scoparius (D, H, L, P) seeds following artificial ageing. The overall volatile composition shows the four major volatiles: ethanol, methanol, acetaldehyde, and acetone, along with the sum of all other volatiles (A–D). The other volatiles are divided into groups: alcohols (E–H); aldehydes (I–L); and miscellaneous compounds (M–P). Each bar represents the mean ±SE of three replicates. Plots with the y-axis shown on the left share the same scale as the left-hand plot within the same row, whilst plots with the y-axis on the right have the same scale as the right-hand plot within the same row. Note that the scale of plot K is multiplied by 10. (This figure is available in colour at JXB online.)

Table 2.

The volatile compounds released by ageing Pisum sativum, Cytisus scoparius, and Lathyrus pratensis seeds, and desiccating Quercus robur seeds were detected using SPME and GC-MS Correlation coefficients (R ²) indicate a significant (P < 0.05) relationship between ageing/desiccation period or germination and individual volatile compounds. Non-significant relationships are indicated by NS.

		Ageing/desiccation period		Germination (%)
		R 2	P-value	R 2	P-value
P. sativum (ageing)	1-[2-Methyl-3-(methylthio)allyl] cyclohex-2-enol	0.733	<0.0001	0.625	<0.0001
	1-Octen-3-ol	0.712	<0.0001	0.642	<0.0001
	1-Pentanol	NS		NS
	1-Propanol	0.535	<0.0001	0.495	<0.0005
	2-Butanone	NS		0.265	<0.05
	2-Butyl-2,7-octadien-1-ol	NS		0.195	<0.05
	2-Methylbutanal	NS		0.228	<0.05
	2-Propanol	0.545	<0.0001	0.406	<0.005
	3-Methylbutanal	NS		NS
	Acetaldehyde	NS		NS
	Acetic acid	0.556	<0.0001	0.604	<0.0001
	Acetone	NS		NS
	Acetonitrile	NS		NS
	Benzaldehyde	NS		NS
	Butanal	NS		NS
	Dimethylsilanediol	NS		NS
	Ethanol	NS		NS
	Ethylbenzene	0.467	<0.0005	0.434	<0.001
	Hexanal	NS		NS
	Methanol	0.603	<0.0001	0.667	<0.0001
	n-Hexylformate	0.276	<0.05	NS
	Pentanal	NS		NS
	Propanal	NS		0.188	<0.05
	Xylene	0.623	<0.0001	0.464	<0.0005
C. scoparius (ageing)	1-Pentanol	NS		NS
	1-Propanol	NS		NS
	1-Propen-2-ol acetate	NS		0.291	<0.05
	2-Butanol	0.318	<0.05	NS
	2-Butyl-2,7-octadien-1-ol	0.261	<0.05	0.281	<0.05
	2-Methyl-1-propanol	NS		NS
	2-Methyl-2-propanol	NS		NS
	2-Propanol	0.548	0.001	NS
	2-Propenenitrile	0.251	<0.05	NS
	3-Carene	0.646	<0.0005	0.324	<0.05
	3-Methylbutanal	NS		NS
	Acetic acid	NS		NS
	Butanal	NS		NS
	Diisobutylene	0.292	<0.05	NS
	d-Limonene	0.571	0.001	0.332	<0.05
	Ethanol	NS		NS
	Ethylbenzene	0.417	<0.01	NS
	Methanol	NS		NS
	n-Hexylformate	NS		NS
	Propanal	NS		NS
	Unknown	0.272	<0.05	NS
	Xylene	0.433	<0.01	NS
L. pratensis (ageing)	1-Octen-3-ol	NS		NS
	1-Pentanol	NS		NS
	1-Penten-3-ol	0.478	<0.05	0.629	<0.005
	2-Butanone	0.476	0.05	0.497	<0.01
	2-Methylbutanal	NS		NS
	3-Methylbutanal	NS		NS
	2-Propanol	NS		NS
	Acetaldehyde	NS		NS
	Acetic acid	0.371	<0.05	0.408	<0.05
	Acetone	NS		NS
	Benzaldehyde	NS		NS
	Ethanol	0.407	<0.05	NS
	Heptanal	NS		NS
	Hexanal	NS		NS
	Methanol	0.474	<0.05	0.499	0.01
	Nonanal	NS		NS
	Pentanal	NS		NS
	Xylene	NS		NS
Q. robur (desiccation)	1-Propanol	0.591	<0.001	0.449	<0.005
	2-Methyl-1-propanol	0.553	<0.001	0.561	<0.001
	2-Methylbutanal	0.282	<0.05	NS
	2-Butanol	0.536	<0.005	0.525	<0.005
	3-Methylbutanal	NS		NS
	Acetaldehyde	0.286	<0.05	NS
	Acetic acid	0.499	<0.005	0.458	<0.005
	Acetone	NS		NS
	Dimethylsilanediol	NS		0.258	<0.05
	Ethanol	0.870	<0.0001	0.698	<0.0001
	Hexanal	NS		NS
	Methanol	0.601	<0.0005	0.553	<0.001
	Pentanal	NS		NS
	Propanal	NS		NS
	Xylene	NS		NS

Assessment of the fatty acid composition of seeds during ageing and occurrence of lipid peroxidation in Pisum sativum seeds

Lipid peroxidation is a well-documented process during seed ageing (Kalpana and Rao, 1996; Aiazzi et al., 1997; Bailly et al., 1998; Goel and Sheoran, 2003; Tammela et al., 2003), and is a potential source of volatile compounds released by ageing seeds. The fatty acid composition of P. sativum, C. scoparius, L. pratensis, and Q. robur seeds was analysed during the respective ageing and desiccation treatments. The fatty acid composition showed little variation in response to ageing or desiccation. The only significant changes were an increase in palmitic acid (P < 0.05) and a decline in the unsaturated:saturated fatty acid ratio (P < 0.01) during ageing of P. sativum seeds, and a decrease in the linolenic acid levels of L. pratensis seeds (Supplementary Table S1 at JXB online). In all four species, linoleic acid, oleic acid, and palmitic acid were the most abundant fatty acids, accounting for almost 90% of the total fatty acid composition. Linoleic acid was the most abundant fatty acid in C. scoparius and L. pratensis seeds, whilst in P. sativum and Q. robur the concentrations of oleic acid and linoleic acid were similar. Linolenic and stearic acid were also present, but at much lower levels.

The products of lipid peroxidation, malondialdehyde and 4-hydroxynonenal, were measured in non-aged P. sativum seeds and seeds aged for 25 d and 55 d. Both compounds increased in seeds aged for 25 d (P < 0.05), and then declined in seeds aged for 55 d (Supplementary Table S2 at JXB online).

Discussion

Artificial ageing of orthodox seeds and desiccation of recalcitrant seeds caused a decline in seed viability, as assessed by germination testing. Seeds that had been aged or desiccated, but were still viable, were slower to germinate, which is considered to be an indicator of loss of vigour. Changes in volatile levels could be detected prior to viability loss. This was particularly clear for P. sativum seeds, where ageing for 8, 12, and 15 d had no effect on total germination, but increases in volatile production were observed even after 8 d of ageing. Analysis of volatiles using SPME showed that the major compounds produced by all seeds were ethanol, methanol, and acetone. Acetaldehyde was also produced in relatively large amounts by all species, with the exception of C. scoparius. These four compounds have been widely reported in other studies of seed volatile production during storage (Zhang et al., 1993, 1995a ; Buckley and Buckley, 2009; Mira et al., 2010). In this investigation, analysis with the Tenax adsorbent showed that ethanol production was inversely correlated with germination for P. sativum and Q. robur seeds following ageing and desiccation, respectively (Table 1). Ethanol was detected in the headspace of all four species when analysed using SPME. However, a relationship with the ageing/desiccation period or viability was only apparent for Q. robur and L. pratensis. Acetaldehyde was not detected using the Tenax adsorbent, but was detected in the SPME analysis of ground P. sativum, L. pratensis, and Q. robur seeds, although it did not show a correlation with viability. Ethanol and acetaldehyde derive from glycolytic reactions, which may still occur at a very slow rate in dry seeds (Zhang et al., 1993). Increased production of ethanol, acetaldehyde, and acetic acid has been reported as a consequence of plant stress (Kreuzweiser et al., 1999). Ethanol is produced under oxygen-limited conditions by alcoholic fermentation in which pyruvate from glycolysis is converted to acetaldehyde by pyruvate decarboxylase. The subsequent conversion of acetaldehyde to ethanol is catalysed by alcohol dehydrogenase. Additionally, acetaldehyde may be oxidized to acetate by aldehyde dehydrogenase (Kreuzweiser et al., 1999). Acetone may also arise from the decarboxylation of pyruvate, as demonstrated by feeding experiments using isotopically labelled sodium pyruvate which resulted in emission of labelled acetone, ethanol, acetic acid, and acetaldehyde from leaves (Jardine et al., 2010). By storing rice or lettuce seeds with gaseous ethanol or acetaldehyde, Zhang et al. (1995b) demonstrated that alcohol dehydrogenase was active even at a RH as low as 12%, and catalysed the interconversion of ethanol and acetaldehyde. In addition, ethyl acetate evolved from lettuce seeds stored at 12% RH with ethanol and acetate, demonstrating that esterification reactions may occur in dry seeds. Therefore, methyl acetate may arise via the esterification of methanol and acetic acid. Alternatively, the transfer of an acetyl group from acetyl-CoA to methanol catalysed by acyl transferases may also give rise to methyl acetate (Pichersky et al., 2006). Whilst not detected using Tenax, methanol was detected in all species using SPME, and showed a positive correlation with the ageing/desiccation period and an inverse correlation with viability for all species except C. scoparius. Methanol may arise from lipid peroxidation, as well as from glycolysis (Mira et al., 2010), and the demethoxylation of pectin by cell wall-associated pectin methyl esterases (Zhang et al., 1993; Kreuzwieser et al., 1999).

In addition to its production during alcoholic fermentation, acetaldehyde is also produced during lipid peroxidation, and the major metabolic source depends on the nature of the biotic or abiotic stress; for example, under oxygen-limiting conditions, alcoholic fermentation is the main source of acetaldehyde (Jardine et al., 2009). Aldehydes are common products of lipid peroxidation, with hexanal and pentanal associated with the degradation of n-6 polyunsaturated fatty acids (e.g. linoleic acid) and propanal with that of n-3 polyunsaturated fatty acids (e.g. α-linolenic acid). Hexanal was detected using SPME in the headspace of all seeds except C. scoparius, and was produced in the greatest abundance by L. pratensis seeds. The levels in P. sativum peaked, along with those of pentanal and propanal, after 12 d of ageing. In P. sativum seeds, the peak and subsequent decline in accumulation of volatile aldehydes is in agreement with the pattern observed for malondialdehyde and 4-hydroxynonenal. Both peroxidation products transiently increased in P. sativum seeds aged for 25 d (P < 0.05) before declining to low levels after 55 d of ageing, possibly due to degradation of malondialdehyde and 4-hydroxynonenal. A similar pattern of malondialdehyde production was observed in aged soybean axes (Stewart and Bewley, 1980). Lipid peroxidation gives rise to a range of other carbonyl compounds such as alkanes, alcohols, ketones, and esters, and is the likely source of many of the alcohols (e.g. propanol, butanol, pentanol, etc.) detected in this study, along with other compounds such as 2-butanone and n-hexylformate. Lipid peroxidation has also been reported as a result of desiccation stress in Q. robur seeds (Finch-Savage et al., 1996), and 2-pentanone, which was detected using Tenax, and hexanal, pentanal, and propanal, which were detected using SPME, are likely to be products of lipid peroxidation caused by desiccation of Q. robur acorns. Decreased unsaturation of fatty acids is also an indicator of lipid peroxidation associated with seed ageing (Stewart and Bewley, 1980; Sung and Jeng, 1994; Sung and Chiu, 1995; Tammela et al., 2005; Walters et al., 2005), and the ratio of unsaturated to saturated fatty acids declined by ~8% (P < 0.01) in P. sativum seeds, whilst the fatty acid composition of the other seeds was not greatly affected by ageing (Supplementary Table S1 at JXB online).

Differences in the volatile profiles may be related to the biochemical composition of the seeds, which tend to be dominated by storage reserves. The seeds used in this study had mainly carbohydrate- and protein-based reserves, with low oil content ranging from 1% in P. sativum to 6% in C. scoparius (Royal Botanic Gardens Kew, 2008). Few studies have reported a relationship between carbohydrate composition and longevity, although changes in reducing sugars have been observed during accelerated ageing and this has been associated with the occurrence of Maillard reactions (Bernal-Lugo and Leopold, 1992). Analyses using Tenax showed greater similarity in the volatile profiles of P. sativum and L. pratensis seeds, with methyl acetate being the major volatile produced, whilst 2-methylbutanal was the predominant volatile released by C. scoparius seeds. 2-Methylbutanal was also produced during seed storage in previous studies (Zhang et al., 1995a ), and was proposed to be a product of the Strecker degradation of isoleucine. Strecker degradation is the oxidative deamination and decarboxylation of α-amino acids, which occurs in the presence of α-dicarbonyl compounds such as 2,3-butanedione formed during Maillard reactions (Cremer and Eichner, 2000). In this study 2,3-butanedione was released by ageing P. sativum and L. pratensis seeds, which provides evidence for the occurrence of Maillard reactions during artificial ageing at 50 °C and 60% RH. Other products of Strecker degradation include acetaldehyde, 2-methylpropanal, and 3-methylbutanal from alanine, valine, and leucine, respectively. All were detected in this study, but there was no consistent relationship with either ageing or viability across all species. Strecker degradation is more likely to occur under conditions of higher RH (e.g. 44–75%), so may not contribute to seed deterioration during storage at low RH and subzero temperatures (Zhang et al., 1995a ). The products of both sugar hydrolysis and lipid peroxidation can initiate non-enzymatic protein and DNA degradation via Amadori and Maillard reactions. Despite reports of the inhibition of enzymatic and auto-oxidation of lipids at the seed MCs used in this study [13.2, 11.4, and 10.2% (fresh mass basis) for P. sativum, L. pratensis, and C. scoparius, respectively; Murthy and Sun, 2000], malondialdehyde and 4-hydroxynonenal were detected in aged P. sativum seeds, indicating that lipid peroxidation did occur and may have contributed to Maillard reactions.

Some unusual compounds were detected using SPME; for example, xylene was found in the headspace of all four species, and xylene levels correlated with ageing of P. sativum seeds. Xylene production has been associated with carotenoid degradation (Rios et al., 2008), which was reported to occur during seed ageing (Pinzino et al., 1999), and may be the source of xylene in this study. Other unusual compounds include acetonitrile, dimethylsilanediol, diisobutylene, and ethylbenzene, but these may be contaminants derived from the vial septa, etc. The potential sources of volatile compounds released by seeds during ageing or desiccation treatments are summarized in Supplementary Fig. S2 at JXB online. The relationship between individual volatiles and ageing or viability was sometimes weak (R ² < 0.5; Table 2; Supplementary Fig. S1), especially for L. pratensis. This may reflect the complexity of studying seed populations in which some seeds are dead, whilst others are viable but in various states of deterioration. In addition, volatiles can derive from more than one source and in some cases may participate in further reactions, so that their release does not necessarily follow a linear pattern.

Comparison of the Tenax adsorbent with SPME revealed that SPME was potentially more sensitive, and more compounds were detected despite exposure of the SPME fibre to a much smaller seed sample and for a shorter period of time. There were differences in the volatile profiles obtained using each technique, which may in part reflect the differences in the experimental approaches; for example, the data obtained using the Tenax adsorbent were based on intact seeds, and the Tenax was exposed to the seeds throughout the ageing or desiccation period, whereas SPME was performed by exposing the SPME fibre to the headspace above freeze-dried seeds, which had been stored at –75 °C following the ageing/desiccation treatments and then ground to a fine powder. The differences may also relate to the selectivity of the adsorbents for different volatiles, and the temperature at which volatile trapping was performed; for example, Tenax was exposed to the seeds during ageing at 50 °C compared with 30°C for the SPME analyses, which will influence the adsorption and release of individual volatiles. Whilst the use of SPME or adsorbents to trap volatile compounds results in much greater sensitivity than direct headspace sampling, inevitably the choice of SPME fibre or adsorbents affects the volatile profiles obtained (Agelopoulos and Pickett, 1998). Therefore, these techniques are ideally suited to comparing the effect of treatments such as ageing on volatile production (so long as all samples are analysed under identical conditions), but are less suited to determining the exact composition of complex volatile mixtures. The advantage of using SPME is that monitoring of seed volatiles is more straightforward without the need to transfer Tenax between storage containers and headspace vials. Manual SPME could be employed for larger seeds.

Conclusions

In this study, the volatile compounds that correlated with seed ageing or desiccation were mainly alcohols, which may derive from glycolytic reactions. Methanol shows the most promise as a seed viability marker because it was produced by all four species, and correlated with viability in all species except C. scoparius, but it could only be detected using SPME analysis. The detection of changes in volatile composition prior to viability loss highlights the potential of volatile analysis as a means of monitoring seed collections for early signs of deterioration. Seed deterioration produced a similar suite of volatiles in both orthodox seeds undergoing artificial ageing and a recalcitrant seed exposed to desiccation stress. This demonstrates that both processes appear to involve disruption of glycolysis, and a switch to alcoholic fermentation. Volatile compounds associated with lipid peroxidation and Maillard reactions were detected, showing that these processes were also occurring, but their contribution to the overall volatile levels was very small in comparison with glycolytic reactions. The chemical processes occurring in seeds are highly dependent on moisture content and temperature, with higher moisture content favouring glycolytic reactions whilst lipid peroxidation tends to occur at lower moisture content (Mira et al., 2010). Further research is needed to elucidate the mechanisms of seed ageing and viability loss during storage, and to determine the effect of seed moisture content and storage temperature on these processes, particularly under dry, low temperature conditions when the cytoplasm is in a glassy state. This study is the first to explore the relationship between volatile production and seed viability under controlled conditions for more than one species. Although there were many similarities between the volatiles evolved by the different species, the pattern and overall composition varied even among the three orthodox species. This highlights the need to conduct a wider investigation to determine the influence of seed storage reserve composition and other inherent factors on seed volatile production.

Supplementary data

Supplementary data are available at JXB online.

Figure S1. Scatter plots of volatile concentration plotted against ageing/desiccation period.

Figure S2. Schematic showing proposed sources of volatile compounds released by seeds.

Table S1. Fatty acid composition of seeds.

Table S2. Malondialdehyde and 4-hydroxynonenal content of P. sativum seeds during ageing.

Glossary

Abbreviations:

GC-MS

gas chromatography–mass spectrometry

LC-APcI-MS

high performance liquid chromatography with atmospheric pressure chemical ionization mass spectrometric detection

MC

moisture content

RH

relative humidity

ROS

reactive oxygen species

SPME

solid phase microextraction.