Longevity of Preserved Germplasm: The Temperature Dependency of Aging Reactions in Glassy Matrices of Dried Fern Spores

Abstract

This study explores the temperature dependency of the aging rate in dry cells over a broad temperature range encompassing the fluid to solid transition (Tg) and well below. Spores from diverse species of eight families of ferns were stored at temperatures ranging from +45°C to approximately –176°C (vapor phase above liquid nitrogen), and viability was monitored periodically for up to 4,300 d (∼12 years). Accompanying measurements using differential scanning calorimetry (DSC) provide insights into structural changes that occur, such as Tg between +45 and –20°C (depending on moisture), and triacylglycerol (TAG) crystallization between –5 and –35°C (depending on species). We detected aging even at cryogenic temperatures, which we consider analogous to unscheduled degradation of pharmaceuticals stored well below Tg caused by a shift in the nature of molecular motions that dominate chemical reactivity. We occasionally observed faster aging of spores stored at –18°C (conventional freezer) compared with 5°C (refrigerator), and linked this with mobility and crystallization within TAGs, which probably influences molecular motion of dried cytoplasm in a narrow temperature range. Temperature dependency of longevity was remarkably similar among diverse fern spores, despite widely disparate aging rates; this provides a powerful tool to predict deterioration of germplasm preserved in the solid state. Future work will increase our understanding of molecular organization and composition contributing to differences in longevity.

Introduction

Germplasm banks aim to maintain viable propagules in perpetuity to serve an ex situ conservation mandate (Guerrant et al. 2004). Potential longevity, in combination with good storage conditions, keeps germplasm alive to meet conservation goals (Hay and Probert 2013). That said, the scientific principles guiding potential longevity as well as good storage conditions typically used in plant germplasm banks, i.e. sub-zero Celsius, are not well understood. This leads to inabilities of germplasm banks to predict how long samples will survive during storage, detect minor symptoms of aging occurring prior to mortality, and effect treatments to postpone or repair aging damage. For example, since the 1990s, the literature is replete with contrasting assumptions about how extremes of temperature, moisture and a gaseous environment during storage affect longevity (Pritchard and Dickie 2003, Walters et al. 2005a, Ellis and Hong 2006, Buitink and Leprince 2008, Groot et al. 2015, Walters 2015). These uncertainties, in turn, prohibit consensus models of aging kinetics among diverse storage conditions.

Drying and cooling cytoplasm leads to long-term physical and chemical stability. The process results in solidification of the cytoplasm, which is prerequisite for germplasm survival if lethal mechanical damage can be avoided as cell constituents compress (Rall and Fahy 1985, Burke 1986). While there is little question that amorphous solids form in the cytoplasm to support preservation (Buitink and Leprince 2008, Walters 2015), regulation of molecular mobility and the inter-relationship with chemical stability remain poorly understood. For example, the role of enzyme activity in relatively dry cytoplasm is still conjectural (Fernández-Marín et al. 2013). Research on stability of foods and pharmaceuticals provides substantial guidance in terms of the nature of deteriorative reactions and their dependence on formulation, temperature and moisture (Yoshioka and Aso 2007, Laitinen et al. 2013). The literature distinguishes between ‘global’ and ‘local’ mobility, which refer to slow, long-distance diffusive motion relevant to glass transitions (also structural relaxation, α relaxation and vitrification↔ plasticization transitions) and fast, short-distance motion (such as ligand rotations and vibrations), also called Johari–Goldstein (β) relaxation (Bhattacharya and Suryanarayanan 2009). Chemical reactivity leading to lost function of materials can be highly dependent on global or local mobility depending on the mechanism of degradative reactions. Bimolecular reactions [such as protein aggregation (Zhou and Labuza 2007) or Maillard reactions (Craig et al. 2001)] are generally affected by global mobility. In contrast, within-molecule interactions are regulated by local mobility [such as deamidation leading to insulin degradation (Yoshioka et al. 2006, Chang and Pikal 2009)]. The rare occasion of a poor correlation between mobility and chemical stability is attributed to structural complexes or heterogeneity within the solidified matrix (Yoshioka and Aso 2007). The diversity of molecules subject to degradation in cytoplasm, as well as wide-ranging constituents that serve as possible antiplasticizers, plasticizers, fillers and carriers that stabilize or destabilize matrix structures, might explain poor correlations between specific molecules and longevity across diverse germplasm and genetic backgrounds (e.g. Hoekstra 2005, Buitink and Leprince 2008, Probert et al. 2009, Birtić et al. 2011, Sano et al. 2016).

Much of the understanding of how molecular mobility correlates with chemical stability is discovered through studies of the temperature dependence or plasticizer effects on deteriorative reactions (Yoshioka and Aso 2007, Bhattacharya and Suryanarayanan 2009, Laitinen et al. 2013). Storage stability can be predicted from accelerated deterioration testing if the reactions involved are regulated by molecular mobility and the temperature dependence of mobility is explicitly described. This is the motivation behind seed longevity models (e.g. Ellis and Roberts 1980) and high moisture and temperature treatments designed to age seeds quickly (e.g. Probert et al. 2009). These models reliably predict longevity at storage conditions above the glass transition (Tg) (Sun 1997, Murthy et al. 2003), but fail when seeds or spores are stored under conditions where global mobility is restricted in the amorphous solid (i.e. below Tg) (Pritchard and Dickie 2003, Walters et al. 2005a, Ellis and Hong 2006, Ballesteros et al. 2017). Current genebanking standards call for storage at conditions well below Tg (FAO 2014). This recommendation ensures that structural relaxations, which become limited at Tg –50°C, are minimal (Shamblin et al. 1999, Walters 2004). However, these storage conditions are ‘uncharted territory’, and we are not aware of any theoretical or data-driven guidelines that help predict how long dried germplasm survives under these conditions.

Quantitative relationships that characterize chemical reactivity and global mobility are based on temperature dependence. Models rely on classic Arrhenius behavior, with potential curvilinear relationships between ln(rate) and the reciprocal of temperature (T^–1) at temperatures near Tg, which tend to fit better to empirically derived Williams–Landel–Ferry (WLF) or Vogel–Fulcher–Tamman (VFT) models (Peleg 2017). The various models differ by the number of parameters defining the slope (1, 2 or 3, respectively) and are otherwise mathematically equivalent (Peleg 2017). The Arrhenius equation [k = A_o·exp(–E_a/RT), where k is rate, A_o is a frequency factor, E_a is energy of activation, R is the ideal gas constant and T is temperature] provides physical significance to the parameters E_a and A_o: E_a (the 1-parameter slope of the Arrhenius plot) is derived from temperature effects on equilibrium constants based on van’t Hoff relationships, and A_o (the pre-exponential constant) is described as the number of ‘diffusional jumps’ needed to effect a diffusion-driven reaction (Chang and Pikal 2009). Classification of ‘fragile’ and ‘strong’ glasses can be based on deviation from linearity near Tg in Arrhenius plots, with strong glasses being more linear and having shallower slopes because they are more resistant to structural changes than fragile glasses (Chang and Pikal 2009).

As temperature declines well below Tg, global mobility (α relaxations) becomes extremely limited and chemical reactivity is low. However, local motions associated with β relaxations continue (e.g. Ballesteros and Walters 2011) and these are increasingly implicated in the remaining chemical reactivity and eventual destabilization (Yoshioka and Aso 2007, Bhattacharya and Suryanarayanan 2009, Chang and Pikal 2009, Laitinen et al. 2013). Sharply contrasting temperature dependencies of degradative reactions below Tg are symptomatic of the conversion from global to local motions (Rault 2000, Yoshioka and Aso 2007, Bhattacharya and Suryanarayanan 2009, Chang and Pikal 2009). A useful analogy is that the same deteriorative effect is accomplished through many tiny, low activation energy steps (Chang and Pikal 2009). Local mobility is detectable in dry seeds (Ballesteros and Walters 2011). Interestingly, the low temperature dependencies appear similar for rotational motion in dried cytoplasm (Buitink et al. 1999) and aging in cryogenically stored seeds (Walters et al. 2004), and contrast with higher temperature dependency measured for diffusional and structural relaxations (Walters 2004). The low temperature dependency of local motions might explain continuation of deteriorative reactions and lower than expected viability in dry, cryogenically stored germplasm (Dickie et al. 1990, Pritchard and Seaton 1993, Ballesteros et al. 2011, Ballesteros and Pence 2017).

Dried cells contain lipid droplets dispersed within the aqueous matrix, and the two domains have very different physical properties. Though there appears to be little interaction of water with triacylglycerols (TAGs), TAGs are still implicated in moisture-regulated deterioration of germplasm (e.g. Hoekstra 2005, Walters et al. 2005b, Lehner et al. 2006), despite poor correlation with lipid content (Probert et al. 2009) and inconsistent change in oxidation status with deterioration (Murthy et al. 2003, Oenel et al. 2017). Several so-called ‘intermediate’ seeds survive drying to a glassy state, but then appear to die more rapidly than expected near freezer temperatures [e.g. papaya (Ellis et al. 1991), some orchids (Pritchard and Seaton 1993), coffee (Eira et al. 2006), cuphea (Crane et al. 2006) and citrus (Hamilton et al. 2009)]. These anomalous temperature effects have been linked to TAG crystallization that occurs below 0°C in fatty acid mixtures containing moderate proportions of medium to long carbon chains having no or one unsaturated bond (Crane et al. 2006, Hamilton et al. 2009).

The goal of this study is to provide better tools to predict longevity of solid-state cytoplasm by connecting principles of molecular mobility with measured stability. We do this by characterizing the kinetics of deterioration in biological systems during storage at temperature and moisture levels at and below Tg. There are numerous examples of experimentally accessible organisms that are desiccation tolerant, and so not susceptible to the problem of lethal ice formation, which would prohibit inferences about amorphous solid cytoplasm at sub-zero temperatures. Moreover, single-celled systems eliminate the confounding effects of differentiated cell structures; structure is an inextricable component of mobility. Single-celled systems also facilitate future opportunities to infuse cells with known matrix stabilizers and destabilizers, an empirical approach that benefits from understanding in foods and pharmaceuticals. Our studies use fern spores, unicellular propagules that appear analogous to seeds in that they tolerate desiccation (Ballesteros et al. 2017) and can remain dormant in the soil to form soil banks (Dyer 1994). There are two types of fern spores: those that retain active Chl in the chloroplasts during maturation and drying (known as green or chlorophyllous spores) and those that do not (known as non-green or non-chlorophyllous spores) (Sundue et al. 2011). Both spore types naturally survive cytoplasmic solidification upon drying, and longevity of dry spores consistently increases with decreasing temperature near Tg (Ballesteros et al. 2017). That said, longevity varies vastly among species, which may suggest differences in the specific chemical reactions that cause aging or the mechanisms that regulate mobility within the solidified matrix. The fern spores used in this work have been stored at a range of temperatures [including those near liquid nitrogen (LN)] and water contents near and below Tg. Storage treatments were begun in 2005, and the multifactorial experimental design provides a unique data set to understand factors that contribute to longevity. We hypothesized that aging kinetics follow similar patterns to diffusion-driven reactions and that temperature dependencies among spores are similar, suggesting commonality in moisture and temperature regulation of molecular motion.

Results

Initial viability and longevity

Initial germination, assessed within either 1 (green spores) or 6 months (non-green spores) after harvest, ranged from 57% to 97% depending on species Supplementary Table S2. Generally, water content adjustment prior to storage did not significantly decrease initial spore germination (binomial test, P > 0.05), except for spores treated at 1% relative humidity (RH) [Culcita macrocarpa (2006H), Polystichum aculeatum, Pteris vittata and Woodwardia radicans] or 75% RH [C. macrocarpa (2006H)] for prolonged periods (binomial test, P < 0.01). Results from these treatments were excluded from subsequent analyses of longevity to ensure that ‘pre-aging’ did not confound interpretation of temperature and moisture interactions.

Table 1

Longevity of fern spores from diverse species stored up to 4,226d (12 years) at different temperatures and moisture conditions

		P50 (days)
Species (population)	RH (%)	45°C	35°C	25°C	5°C	–18°C	–80°C	–176°C (LN)
Culcita macrocarpa (2006H)	11–15	0.6 ± 0.5	14 ± 1	29 ± 1	478 ± 21	503 ± 51	993 ± 195	698 ± 599
		(0.1)	(16)	(40)	(552)	(1,081)	(1,150)	(1,036)
	47–61	2.5 ± 0.3	28 ± 1	50 ± 2	393 ± 31	694 ± 22	1,778 ± 21	2,321 ± 115
		(0.6)	(20)	(52)	(755)	(663)	(936)	(870)
Culcita macrocarpa	11–15	1 ± 0.2	9 ± 1	23 ± 1	799 ± 35	1,900 ± 84	1,575 ± 66	3,513 ± 463*
(Galicia)		(1.6)	(6.5)	(39)	(799)	(1,936)	(1,943)	(2,732*)
	31–33	0.3 ± 0.02	9.3 ± 0.5	19 ± 1	839 ± 55	N/A	N/A	N/A
		(0.1)	(4.9)	(29)	(955)	N/A	N/A	N/A
	47–61	0.9 ± 0.1	0.6 ± 0.1	13 ± 0.5	62 ± 2	N/A	N/A	N/A
		(1.5)	(3)	(50)	(101)	N/A	N/A	N/A
	75a	N/A	N/A	8.4 ± 0.4	17 ± 4	N/A	N/A	N/A
		N/A	N/A	(15)	(61)	N/A	N/A	N/A
Culcita macrocarpa	11–15	9 ± 0.3	36 ± 1	46 ± 2	584 ± 19	1,004 ± 24	1,694 ± 92	1,709 ± 64*
(Azores)		(11)	(48)	(129)	(977)	(1,436)	(1,884)	(1,923*)
	75a	N/A	N/A	21 ± 1	95 ± 4	N/A	N/A	N/A
		N/A	N/A	(27)	(41)	N/A	N/A	N/A
Dicksonia antarctica	11–15	61 ± 1	115 ± 1	465 ± 10	3,778 ± 318*	2,071 ± 34*	4,941 ± 469*	4,853 ± 439*
		(59)	(131)	(408)a	(4,725*)	(2,058*)	(117,494*)	(234,106*)
	31–33	9.7 ± 0.2	102 ± 1	175 ± 2	3,181 ± 210*	N/A	N/A	N/A
		(5.7)	(108)	(144)a	(6,271*)	N/A	N/A	N/A
	47–61	6 ± 0.1	14 ± 0.1	143 ± 1	535 ± 284*	N/A	N/A	N/A
		(5)	(16)	(128)a	(930*)	N/A	N/A	N/A
	75a	N/A	N/A	25 ± 1	394 ± 14	N/A	N/A	N/A
		N/A	N/A	(21)	(264)	N/A	N/A	N/A
Equisetum hyemale	11–15	2 ± 0.1	6 ± 0.1	11 ± 0.2	93 ± 2	907 ± 46	1,117 ± 40	753 ± 41
(2010H)		(1.7)	(3.7)	(12)	(94)	(790)	(1,974)	(1,797)
	31–33	0.8 ± 0.1	3.6 ± 0.1	5.9 ± 0.1	67 ± 1	549 ± 15	864 ± 23	1,210 ± 27
		(0.4)	(3.2)	(6)	(78)	(743)	(1,174)	(1,660)
	47–61	0.8 ± 0.1	3.4 ± 0.1	7.3 ± 0.1	80 ± 1	610 ± 18	1,218 ± 31	1,293 ± 21
		(0.3)	(3.3)	(8)	(88)	(815)	(1,847)	(1,601)
	75a	N/A	N/A	2.6 ± 0.1	37 ± 1	N/A	N/A	N/A
		N/A	N/A	(1.9)	(43)	N/A	N/A	N/A
Equisetum hyemale	11–15	N/A	N/A	N/A	66 ± 3	612 ± 18	1,295 ± 1,671	1,328 ± 4,734
(2013H)		N/A	N/A	N/A	(76)	(575)	(1,009)	(1,059)
	31–33	N/A	N/A	N/A	59 ± 1	455 ± 27	1,306 ± 2,077	1,313 ± 2,489
		N/A	N/A	N/A	(69)	(504)	(1,024)	(1,036)
	47–61	N/A	N/A	N/A	70 ± 1	482 ± 32	N/A	N/A
		N/A	N/A	N/A	(82)	(535)	N/A	N/A
Matteuccia struthiopteris	11–15	19 ± 0.3	56 ± 0.6	119 ± 1	994 ± 8	5,610 ± 486*	7,240 ± 978*	5,815 ± 769*
(BH)		(19)	(52)	(131)	(1,246)	(2,993*)	(4,152*)	(2,844*)
	75	1 ± 1	3 ± 0	24 ± 0.3	119 ± 5	6,223 ± 674*	5,579 ± 527*	6,082 ± 943*
		(1)	(6)	(25)	(264)	(3,346*)	(8,722*)	(3,342*)
Matteuccia struthiopteris	11–15	N/A	N/A	N/A	837 ± 8	3,020 ± 153*	3,380 ± 158*	4,829 ± 359*
(DBG)		N/A	N/A	N/A	(909)	(2,268*)	(3,074*)	(3,911*)
Osmunda regalis	11–15	N/A	N/A	92 ± 4	310 ± 13	880 ± 49*	–b	–b
		N/A	N/A	(118)	(303)	(2,369*)	(–b)	(–b)
	31–33	N/A	N/A	39 ± 1	350 ± 12	769 ± 142*	–b	1,094 ± 135*
		N/A	N/A	(30)	(318)	(3,228*)	(–b)	(–b)
	47–61	N/A	N/A	31 ± 1	230 ± 10	994 ± 79*	–b	–b
		N/A	N/A	(29)	(217)	(2,011*)	(–b)	(–b)
	75a	N/A	N/A	10 ± 0.3	70 ± 1	N/A	N/A	N/A
		N/A	N/A	(5)	(64)	N/A	N/A	N/A
Polystichum aculeatum	11–15	77 ± 3	386 ± 18	872 ± 30	9,105 ± 2,244*	3,606 ± 149*	–b	12,200 ± 3,541*
		(64)	(430)	(852)	(12,822*)	(16,473*)	(–b)	(9 × 106*)
	47–61	33 ± 1	347 ± 13	993 ± 29	7,732 ± 979*	5,955 ± 349*	8,440 ± 1,335*	16,239 ± 6,583*
		(22)	(455)	(852)	(7,962*)	(16,134*)	(3 × 109*)	(4 × 1017*)
	75	N/A	N/A	N/A	N/A	3,692 ± 116	7,581 ± 690*	17,442 ± 7,385*
		N/A	N/A	N/A	N/A	(2,041)	(1 × 108*)	(3 × 106*)
Pteris vittata	11–15	411 ± 16	2,851 ± 274*	4,781 ± 199*	–b	6,378 ± 541*	–b	7,736 ± 961*
		(426)	(2,391*)	(4,792*)	(125,774*)	(11,110*)	(6 × 1015*)	(2 × 106*)
	47–61	62 ± 1	4,569 ± 1,278*	4,650 ± 180*	9,950 ± 1,579*	6,781 ± 598*	8,246 ± 1,143*	5,865 ± 381*
		(65)	(3,145*)	(4,956*)	(5,163*)	(4,885*)	(14,679*)	(28,684*)
	75	N/A	N/A	N/A	N/A	6,762 ± 670	7,082 ± 532*	5,113 ± 240*
		N/A	N/A	N/A	N/A	(8,900*)	(8,921*)	(6,268*)
Woodwardia radicans	11–15	27 ± 1	64 ± 1	126 ± 3	3,151 ± 90	6,146 ± 440*	5,661 ± 362*	5,408 ± 210*
		(28)	(70)	(140)	(3,445)	(5,853*)	(4,369*)	(5,635*)
	47–61	25 ± 1	77 ± 1	105 ± 3	1,841 ± 49	3,023 ± 55	5,596 ± 336*	4,232 ± 88*
		(30)	(80)	(115)	(1,883)	(4,318)	(4,928*)	(4,381*)
	75	N/A	N/A	N/A	N/A	3,128 ± 123	6,091 ± 325*	5,339 ± 174*
		N/A	N/A	N/A	N/A	(2,802)	(4,818*)	(5,512*)

Longevity was measured as time for initial germination to decrease by 50% (P50, in days) and was calculated by fitting data to a logistic regression or the Avrami equation (in parentheses).

Relative humidity (RH) ranges were achieved either by storing the spores over various saturated salt solutions or by sealing spores in DSC pans after moisture adjustments over various saturated salt solutions for 5–7 d at 20 ± 2°C (Ballesteros et al. 2017).

Numbers following the ± represent the SE of the estimate of the logistic regression, and the given range is the 85% confidence interval for the calculated P50.

^a Values published in Ballesteros et al. (2017).

^b Insufficient storage time to observe viability decline to fit aging models; P50 not calculated, but > maximum storage time listed in Supplementary Table S2.

*P50s extrapolated from Avrami equations or from logistic curve for dose (time) response to calculate time for 50% viability decrease.

N/A: data non-available; insufficient spores to include this treatment.

Fern spore viability decreased during 4,300 d (∼12 years) of storage (Figs. 1–3). Time courses revealed an initial phase marked by experimental variation but no apparent change in viability, followed by a phase of abrupt and consistent decline, the latter phase indicating typical aging symptoms and encompassing the time to 50% of initial germination (P50) (Table 1). Most spores stored at 11–14% RH and 25°C died within 365–1,900 d (1–5 y), with the exception of Equisetum hyemale, which died within 1 month (Fig. 3A), C. macrocarpa which died within 3–6 months (Fig. 2) and P. vittata which is still viable after 4,300 d (∼12 years) (Fig. 1C). Deterioration rates varied among harvest years and populations (an example of three accessions of C. macrocarpa harvested from two locations in two different years is provided in Fig. 2). Deterioration was rapid for spores stored at 35 and 45°C (deterioration time courses for P. vittata are presented in Fig. 1C) and slow for spores stored at ≤5°C (Figs. 1–3).

Table 2

Results of regression between aging rate [ln(P50^–1)] and T^–1 (K) (Arrhenius plots) of fern spores stored at various moisture conditions and temperatures ranging from 45 to –176°C

Species (accession)	Moisture (% RH)	Temperature ≥5°Ca				Temperature ≤18°Cb				Temperature break (°C)
		Slope	R2	F-value	Ea (–kJ mol–1)	Slope	R2	F-value	Ea (–kJ mol–1)
Culcita macrocarpa	11–15	–13.6a	0.93	0.038	113.1	–0.07a	0.28	NS	0.6	4.8
(Galicia 1)	47–61	–10.3a	0.92	0.042	85.4	–0.15a	0.64	NS	1.2	0.4
Culcita macrocarpa	11–15	–14.2a	0.99	0.007	118.3	–0.11	0.85	NS	0.9	1.5
(Galicia 2)	31–33	–16.8b	0.99	0.003	139.9	–	–	–	–	–
	47–61	–9.8a	0.95	0.026	81.6	–	–	–	–	–
	75	–3.0a	–	–	24.8	–	–	–	–	–
Culcita macrocarpa	11–15	–8.8a	0.97	0.014	73.3	–0.05	0.43	NS	0.5	–1.9
(Azores)	75	–6.2a	–	–	51.2	–	–	–	–	–
Dicksonia antarctica	11–15	–9.4a	0.99	0.004	77.9	–0.03	0.93	NS	0.3	4.9
	31–33	–12.0a	0.96	0.020	99.9	–	–	–	–	–
	47–61	–10.2a	0.93	0.038	84.8	–	–	–	–	–
	75	–11.4a	–	–	94.5	–	–	–	–	–
Equisetum hyemale	11–15	–8.5a	0.99	0.002	70.9	0.04a	0.56	NS	–0.3	–15
(2010)	31–33	–9.4a	0.98	0.010	77.9	–0.09a	0.82	NS	0.8	–12.3
	47–61	–9.9a	0.99	0.005	82.0	–0.08a	0.48	NS	0.7	–11.6
	75	–10.9a	–	–	90.7	–	–	–	–	–
Matteuccia struthiopteris	11–15	–8.6a	0.99	0.001	71.5	0.01a	0.05	NS	–0.1	–10.7
(Black Hills)	75	–10.7a	0.95	0.025	89.1	–0.002a	0.02	NS	0	–18.9
Osmunda regalis	11–15	–5.4a	–	–	45.0	–	–	–	0.6	–0.6
	31–33	–9.1a	–	–	75.6	–	–	–	4.6	–0.5
	47–61	–9.0a	–	–	75.0	–	–	–	0.7	–3.7
	75	–8.3a	–	–	68.7	–	–	–	–	–
Polystichum aculeatum	11–15	–10.2a	0.99	0.007	84.5	–0.11a	0.43	NS	0.9	9.3
	47–61	–11.3a	0.94	0.032	94.1	–0.15a	0.97	0.026	1.2	8.8
	75	–	–	–	–	–0.22a	0.92	NS	1.8	–
Pteris vittata	11–15	–11.5a	0.88	NS	96.0	–0.03a	–	–	0.3	24.8
	47–61	–9.1a	0.58	NS	75.4	0.04a	0.49	NS	–0.4	13.6
	75	–	–	–	–	0.05a	0.90	NS	–0.4	–
Woodwardia radicans	11–15	–10.6a	0.98	0.012	87.8	0.02a	0.77	NS	–0.1	–1
	47–61	–9.3a	0.97	0.016	77.1	–0.02a	0.06	NS	0.2	–0.6
	75	–	–	–	–	–0.05a	0.25	NS	0.4	–

Lines are calculated using regressions for temperature data ≥5°C and ≤–18°C (Figs. 4–6) (R², F-value provided for lines using ≥3 data points).

Different letter superscripts next to slope values indicate significant differences (P < 0.05).

Activation energy (E_a) is calculated for each line and the temperature at which lines intersect (Arrhenius break) is indicated (last column).

^aLine 1 was calculated for viability loss between 45 and 5°C (most treatments) or 25 and 5°C (for RH = 75% [D. antarctica, C. macrocarpa (Galicia 2), C. macrocarpa (Azores)], and E. hyemale and O. regalis (all moisture treatments).

^bLine 2 was calculated for viability loss between –18 and –176°C. The –18°C point was excluded for D. antarctica and P. aculeatum as spores stored at –18°C aged faster than spores stored at 5°C (‘–18°C anomaly’). In these specific cases, slopes were calculated using data from 5, –80 and –176°C storage.

Fig. 1

Change in germination percentage of spores dried to 13% relative humidity (RH) and stored for up to 4,300 d (∼12 years). Depicted are time courses obtained for non-green spores from diverse species harvested in 2010 (A) or 2005–2006 (B–D). Time courses for higher RH treatments are of a similar shape but indicate a steeper decline. Symbols represent storage temperature: –176°C (open circles), –80°C (filled circles), –18°C (filled triangles), 5°C (open diamonds), 25°C (open squares), 35°C (filled squares) and 45°C (asterisks). Aging time courses at 35 and 45°C are only depicted for Pteris vittata to accommodate the broad time scale of the figure. Curves were drawn by fitting data to the Avrami model: solid curves for actual germination values and dashed curves for projected germination. The dashed horizontal line is 50% of maximum germination used to calculate P50. Shaded vertical arrows on the x-axis indicate values of P50, and P50 values calculated for all treatments are summarized in Table 1.

Fig. 2

Change in germination percentage of Culcita macropcarpa (non-green) spores dried to 13% relative humidity and stored for up to 4,300 d (∼12 years). Depicted are time courses obtained for the same population of spores harvested in two different years (A, B) or from different populations within the same year (B, C). Symbols and curves are as indicated in Fig. 1. Calculated P50 values for all treatments are summarized in Table 1.

Fig. 3

Change in germination percentage of spores dried to 13% relative humidity and stored for up to 2,590 d (∼7 years). Depicted are time courses obtained for green spores from diverse species harvested in 2010 (A), 2009 (B) or 2016 (C). Symbols and curves are as indicated in Fig. 1. Calculated P50 values for all treatments are summarized in Table 1.

For simplicity, we quantified longevity as time to 50% of initial germination (P50) (Table 1). P50 values, calculated separately using either Avrami or logistic functions, appear poorly but significantly correlated for all treatments (r² = 0.14, P = 0.043), and this reflects the high uncertainty of longevity in spores showing no or minor symptoms of deterioration, such as P. vittata, P. aculeatum and Dicksonia antarctica stored at ≤5°C for approximately 4,300 d (∼12 years), and Osmunda regalis stored at ≤–80°C for 675 d (almost 2 years). Omitting 17 (out of 146 treatments) of the longest lived treatments from correlations of P50s calculated using Avrami or logistic functions gave a highly significant relationship (y = 1.02x + 42.3, r² = 0.83, P = 0.002). The high uncertainty for long-lived materials was mitigated by correlating the reciprocal of longevity (aging rate = P50^–1), and P50^–1 values calculated using Avrami or logistic functions were highly correlated for all 146 treatments (y = 0.22x + 0.052, r² = 0.49, P = 0.007). This mathematical transformation increased uncertainty of P50^–1 for very short-lived spores which reached 0% germination within a few days and before the first monitoring interval (i.e. E. hyemale and Matteuccia struthiopteris stored at 45°C and C. macrocaprpa stored at ≥35°C). Omitting nine (out of 146 treatments) shorter lived treatments increased the significance of the correlation between aging rates calculated using Avrami or logistic functions (y = 0.94x + 0.002, r² = 0.88, P = 0.001).

Spores stored at –80 and –176°C showed progressive loss of viability with time for most treatments [general linear model (GLM) P < 0.01]. There are a few cases in which no loss in viability was measured, for example P. aculeatum and P. vittata [after 4,217 d (∼12 years)] and O. regalis [after 675 d (1.9 years)] [Table 1, treatments marked by superscript^b did not change significantly with time; Fig. 4C, data for P. vittata spores stored at –80°C 14% RH (P > 0.05)]. Illustrating the extreme case of rapid degradation under cryogenic conditions is E. hyemale (2010 harvest) spores, which lost all viability within 2,590 d (∼7 years) for all moisture and temperature combinations (Fig. 3A). Longevity differences of spores stored at –80 and –176°C were minor and relatively inconsistent among samples; P50 values rarely even doubled in LN-stored samples despite a 90°C decrease in storage temperature (Table 1).

Fig. 4

Germination percentage of fern spores measured in 2017. Spores were dried to the indicated relative humidity (RH), hermetically sealed and stored at the indicated temperature until harvest. The horizontal solid and dotted lines represent the initial germination and SD (i.e. maximum germination). The 13% RH treatments are also provided as the final point in Figs. 1 or 3, and are provided here for comparison with initial and other RH treatments. Different letters indicate statistically significant different germination among RH treatments at each temperature (P < 0.05 in binomial test). Asterisks indicate statistically significant different germination compared with the initial level (P < 0.05 in binomial test).

Moisture had a large and detrimental effect on P50 in samples stored at temperatures ≥5°C [Table 1; representative time courses provided in Ballesteros et al. (2017)]. We did not observe a consistent pattern of higher moisture and reduced longevity at temperatures ≤–18°C, as indicated by similar P50 values among all moisture conditions (Table 1) (moisture was not a significant factor (P > 0.05) in GLMs of storage time, at temperature ≤–18°C except for P. aculeatum and W. radicans stored at –18°C for which the germination level was significantly lower in spores treated at 75% RH compared with lower RH (P < 0.05)).

Temperature dependency of longevity

Temperature had profound effects on aging rate (P50^–1) for storage treatments between 45 and 5°C [analysis of covariance (ANCOVA); F = 30.2, df = 1, P < 0.001), which were characterized using Arrhenius plots (Figs. 5,6). An apparent linear relationship was observed (Fig. 5A) and correlation coefficients usually exceeded 0.9 [r² ranged from 0.58 (P. vittata, 55% RH) to 0.99 (Table 2)]. Slopes ranged from –3 to –17 and averaged –9.9 ± 2.6, corresponding to an apparent activation energy (E_a) of 82 ± 22 kJ mol^–1 (Table 2). At this temperature range, Arrhenius slopes did not differ significantly among species, except for O. regalis and C. macrocarpa (Galicia 2) at RH approximately 13% (ANCOVA, P = 0.013, Fig. 5A) and for C. macrocarpa (Galicia 2) at RH approximately 33% (ANCOVA, P = 0.016) which had the highest Arrhenius slope among all treatments. Also, Arrhenius slopes appeared unaffected by moisture level within each species (P > 0.05; Table 2; Fig. 5B) with the exception of C. macrocarpa (Galicia 2) in which Arrhenius slope appeared to decrease with increasing moisture (y = 0.202x – 19.58, R² = 0.77, F = 0.128)

Table 3

Melting behavior of TAG within fern spores stored for different periods at –20°C

	Cooled at 10°C min–1 (conventional method)			Annealed at –20°C for 1 h			Stored at –20°C for 5–14 d			Stored at –20°C for 1,900–4,300 d (5–12 years) (except O. regalis)
Species	Enthalpy (J g-1 DW)	No. of peaks	Peak temperature (°C)	Enthalpy (J g–1 DW)	No. of peaks	Peak temperature (°C)	Enthalpy (J g–1 DW)	No. of peaks	Peak temperature (°C)	Enthalpy (J g–1 DW)	No. of pPeaks	Peak temperature (°C)
C. macrocarpa (2006H)	7.9 (7.2)	2.5	–27.0	9.5 (5.8)	3.8	–12.9	10.1 (2.7)	5.0	–12.4	32.0 (0.6)	3.5	–11.0
C. macrocarpa (Galicia)	21.3 (3.3)	3.0	–17.2	31.6 (5.1)	4.1	–11.0	34.9 (5.5)	4.0	–10.7	37.8 (2.8)	4.5	–14.0
C. macrocarpa (Azores)	26.1 (3.9)	3.0	–14.0	40.5 (4.3)	3.6	–9.3	46.9 (6.3)	4.0	–12.9	48.5 (8.5)	4.0	–12.4
D. antarctica	22.0 (2.7)	1.5	–20.3	23.2 (4.9)	3.5	–13.8	33.8 (2.4)	5.0	–9.9	31.3 (0.2)	3.0	–13.1
E. hyemale (2010)	1.4 (0.7)	2.0	–5.7	1.7 (1.0)	1.9	–4.5	4.6 (0.9)	3.5	–2.0	4.5 (2.0)	2.4	–5.5
M. struthiopteris	0.3 (0.3)	1.0	–36.1	0.5 (0.4)	1.3	–27.2	1.8 (0.9)	0.4	0.2	0.5 (0.3)	1.5	–35.9
O. regalis	1.7 (0.5)	1.5	–17.1	3.0 (0.6)	1.4	–4.1	4.9 (1.6)	2.1	–3.5	5.5 (0.9)a	1.6	–2.7
P. aculeatum	23.5 (2.7)	1.6	–22.3	21.8 (2.5)	1.9	–21.1	31.2 (7.3)	3.0	–13.1	38.6 (4.3)	2.3	–10.7
P. vittata	24.2 (7.7)	2.2	–8.6	28.1 (11.0)	2.1	–3.2	31.1 (14.8)	4.0	–6.1	48.0 (26.0)	3.0	–5.6
W. radicans	4.4 (2.8)	2.9	–32.2	6.2 (4.6)	2.1	–18.6	11.3 (9.7)	3.6	–10.5	25.4 (8.5)	2.6	–18.4
Min	0.3	1	–36.1	0.5	1.3	–31	1.8	2.2	–13.1	0.5	1	–35.9
Max	26.1	3	–5.7	40.5	4	–3.3	46.9	5	–3.4	48.5	4.5	–2.7
Median	14.6	2.1	–18.8	15.6	2.2	–12.5	21.2	4	–10.8	31.6	2.8	–11.7
Mean	13.3			16.6			21.1			27.2
SD	11.0			14.2			16.2			17.8

Values represent average enthalpy (SD) of at least five replicates for each species and storage time.

Representative DSC scans are provided in Fig. 7.

^aSample stored for 675 d (1.8 years) at –20°C.

Fig. 5

Arrhenius plots showing the temperature dependency of the aging rate of spores from diverse fern species that have been stored at diverse relative humidity (RH) and temperatures ≥5°C. Aging rate is expressed as P50^–1 (Table 1). Arrhenius plots are given for spores from diverse species (A) or spores treated at different RHs (B). Culcita macrocarpa (Galicia 2), CM; Dicksonia antarctica, DA; Pteris vittata, PV; Equisetum hyemale, EH; Matteuccia struthiopteris, MS. Open symbols and dotted lines represent green spores, and filled symbols and solid lines represent non-green spores. Black symbols and lines represent dry (_d) spores (stored at 11–15% RH), and gray symbols and lines represent humid (_w) spores (stored at 75% RH). Slopes of all linear regressions among species and between RH conditions are not significantly different (Table 2), except for C. macrocarpa (which is represented by asterisks and a dashed line).

Fig. 6

Arrhenius plots showing the temperature dependency of the aging rate of spores from diverse fern species that were dried to 13% relative humidity and stored at temperatures between 45 and –176°C. Aging rate is expressed as P50^–1 (Table 1). Symbols and curves are as indicated in Fig. 5, and results of linear regression calculations are summarized in Table 2. The intersection of lines indicates the break in Arrhenius behavior (temperatures listed in Table 2).

Arrhenius plots of aging rates at temperatures ≤–18°C) contrast sharply with the ≥5°C observations (Fig. 6; Table 2). Below 5°C, Arrhenius slopes were significantly shallower by about two orders of magnitude [see values in Table 2; slopes above and below 5°C are significantly different (P < 0.05)]. Underestimated longevity at –80 and –176°C might explain the insensitivity of the aging rate to low temperature; nevertheless, the significant loss in viability for most samples (reported above) suggests that our longevity estimates are realistic. Moisture had no effect on E_a within this temperature range (df = 3, P-value = 0.481).

The –18°C anomaly

The previous paragraphs describe sharply different effects of temperature on aging rate at ≥5 and ≤–18°C, which points to a discontinuity within the 5 to –18°C temperature range that is explicitly calculated from the intersection of the two Arrhenius plots (Table 2). Independent of this observation, we also observed apparently faster aging of some non-green spores stored at –18°C compared with counterparts stored at 5°C (comparison 95% confidence intervals of P50 for P. aculeatum and P. vittata stored at RH ∼13% and ∼50%, and D. antarctica stored at RH ∼13%). This so-called –18°C anomaly is also observed as lower germination proportions in freezer- (–18°C) compared with refrigerator- (5°C) stored seeds, characteristic of the so-called intermediate seed storage category (Ellis et al. 1991, Walters 2015). In our hands, damage to fern spores during –18°C storage was not measured immediately; for example, it took >1,000 d to detect the difference in D. antarctica (Fig. 1A).

Differential scanning calorimetry (DSC)

The largest structural transitions in fern spores are attributed to crystallization and melting of lipids, presumably TAGs. These transitions were observed in all fern spores at temperatures near the Arrhenius break noted for spore aging, but the complexity, temperature and total enthalpy varied among species and thermal history. During cooling at 10°C min^–1 (conventional DSC protocols in the literature), the average temperature of the first exothermic peak, indicative of the onset of TAG crystallization, was –22.8 ± 5.5°C among all species studied Supplementary Table S3 with TAGs in E. hyemale and P. aculeatum crystallizing at the warmest and coldest temperatures, respectively. Melting transitions, measured following 10°C min^–1 cooling and subsequent rewarming, occurred over a broad temperature range, beginning near –50 to –35°C and ending between –5 and +20°C. TAGs from E. hyemale and P. vittata spores melted at higher temperatures (>-9°C), while TAG in other species had large melting peaks between –25 and –15°C or lower (Figs. 7,8). We characterized the number of peaks and temperature of the most prominent peak as a way to track changes in the polymorphic crystalline structure of TAG. Each species had a characteristic endotherm shape and size, with 1–3 peaks or shoulders and average total enthalpy of 13.3 ± 11 J g^–1 DW across species when spores were cooled and warmed at 10°C min^–1. Enthalpy was lower in green spores, ranging from 0.3 (M. struthiopteris) to 1.7 (O. regalis) J g^–1 DW compared with 4.3 (W. radicans) to 26.1 [C. macrocarpa (Azores)] J g^–1 DW for non-green spores (Table 3). Enthalpy of TAG crystallization was usually less than for melting in non-green spores, and this difference correlated with the onset temperature for crystallization (R² = 0.45; F = 8.45, n = 10, P = 0.015). Lower enthalpies of crystallization compared with melting can be indicative of slow crystal growth that went undetected during 10°C min^–1 cooling.

Fig. 7

The effect of storage time at –20°C on melting characteristics of TAGs within fern spores. Differential scanning calorimetry (DSC) thermograms of diverse fern spores are presented and are representative of other species studied but not presented. Programmed cooling was used for samples cooled at 10°C min^–1 (see Supplementary Table S3 for characterization) or annealed for 60 min; otherwise, spores were stored for the indicated time in a mechanical freezer or in the freezer vault at NLGRP, plunged into LN and loaded into the calorimeter pre-cooled to –170°C (see the Materials and Methods). Samples cooled at 300°C min^–1 were samples that had been stored at –80 or –176°C, rewarmed, plunged into LN and loaded into the calorimeter pre-cooled to –170°C. All samples were heated at 10°C min^–1 in scanning mode, and the endothermic direction is upwards. Power values from DSC were corrected for sample mass, which was usually between 0.5 and 2 mg. Thermogram parameters are summarized in Table 3.

Fig. 8

The effect of storage temperature on lipid melting characteristics within fern spores. Differential scanning calorimetry (DSC) thermograms of diverse fern spores stored for 1,900–4,300 d (5–12 years) at the indicated temperatures are presented (and are representative of other species studied but not presented). Samples, which were sealed into DSC pans soon after harvest, were plunged into LN and loaded into the calorimeter pre-cooled to –170°C (see the Materials and Methods). All samples were heated at 10°C min^–1 in scanning mode and the endothermic direction is upwards. Power values from DSC were corrected for sample mass, which was usually between 0.5 and 2 mg, except for Matteuccia strupthiopteris stored at 5°C (D), for which there was only 0.2 mg of sample. Despite the poor quality of the scan of M. strupthiopteris spores stored at 5°C, it is possible to distinguish TAG transitions similar to the material stored at –20°C, but with lower enthalpy. Thermograms of freshly harvested spores were made in the indicated year and serve as a basis for comparison for the progress of TAG crystallization during storage. Thermogram parameters are summarized in Table 4.

The size, shape and temperature of TAG endothermic events were highly affected by thermal history, especially time at sub-zero temperature. Transition enthalpy increased with increasing time at –20°C [regression of storage time with enthalpy normalized to 10°C min^–1 for non-green spores up to 4,300 d (∼12 years): slope = 0.15, r² = 0.45, F = 21.4, df = 26, P < <0.01; for green spores up to 14 d: slope = 159, r² = 0.77, F = 23.8, df = 7, P < <0.01]. Peak temperature also increased with storage time. For example, storing D. antarctica spores at –20°C progressively shifted melting peaks to higher temperatures and increased melting enthalpy (Fig. 7C; Table 3). The number of peaks appeared to increase with storage time over days, but decreased after years of storage (Fig. 7; Table 3).

Characteristics of TAG melting transitions were also affected by storage temperature. Shape, size and temperature of endotherms from spores cold-loaded from –176°C storage were most similar to thermograms of freshly harvested spores (bottom scan in each panel of Fig. 8). In contrast, enthalpy of TAG melting transitions was usually greater for spores cold-loaded from –80°C or –18°C storage, compared with the fresh sample scanned years ago or counterparts stored for years at –176 or 5°C [paired t-test, P < 0.02 n = 8 (5°C) or 10 (–176°C)]. Melting enthalpy of –80°C spores tended to be larger than that of counterparts stored at –18°C, except for spores of C. macrocarpa and D. antarctica (Table 4, paired t-test, P = 0.10, n = 10). The long-term effect of storage on TAG melting characteristics was reversed by warming the sample to melt TAGs, and samples stored at –80 and –176°C, and sometimes at –18°C, that were recooled and rewarmed had characteristics that were indistinguishable from those of the freshly harvested sample (compare for P. aculeatum in Figs. 7A and 8A; other data not presented). Collectively, these thermograms indicate progressive crystallization of TAGs during storage at –18 and –80°C.

Table 4

Melting behavior of TAGs within fern spores stored for 1,900–4,300d (5–12 years) at different temperatures (see the Materials and Methods)

		Enthalpy of TAG melt in cold loaded samples (J g–1 DW)				Temperature of largest peak in cold-loaded samples (°C)
Species	Years stored	5°C	–20°Ca	–80°C	–176°C	5°C	–20°Ca	–80°C	–176°C
Culcita macrocarpa (2006H)	11.6	7.1 (8.8)	32.0 (0.6)	26.6 (1.1)	15.6 (0.2)	–20.6	–11.0	0.6	–20.4
Culcita macrocarpa (Galicia)	5.3	No sample	37.8 (2.8)	38.0 (7.9)	18.5 (8.3)	–	–14.0	–10.9	–16.2
Culcita macrocarpa (Azores)	5.2	3.7 (0.9)	48.5 (8.5)	44.5 (14.2)	30.3 (8.2)	–11.1	–12.4	–11.5	–32.1
Dicksonia antarctica	6.5	21.3 (4.2)	31.3 (0.2)	23.9 (8.4)	20.6 (0.3)	–23.1	–13.1	–14.4	–17.1
Equisetum hyemale (2010)	7.1	No sample	4.5 (2.0)	10.1(5.5)	12.2 (11.3)	–	–5.5	–7.2	3.7
Matteuccia struthiopteris	6.5	0.3 (0.4)	0.5 (0.3)	3.2 (2.9)	1.3 (0.2)	–26.8	–35.9	–23.1	–22.8
Osmunda regalis	1.8	1.6 (0.8)	5.5 (0.9)	5.2 (1.9)	2.7 (1.9)	–11.2	–2.7	–7.1	–7.1
Polystichum aculeatum	11.6	22.4 (14.4)	38.6 (4.3)	59.8 (20.6)	25.1 (9.3)	–10.5	–10.7	–12.1	–19.9
Pteris vittata	11.6	13.5 (2.0)	48.0 (26.0)	88.2 (23.8)	23.1 (1.2)	–4.5	–5.6	2.1	–5.7
Woodwardia radicans	11.6	4.9 (3.2)	25.4 (8.5)	35.5 (23.8)	30.3 (29.3)	–47.6	–18.4	–20.7	–17.4
Min		0.3	0.5	3.2	1.3	–47.6	–35.9	–23.1	–32.1
Max		22.4	48.5	88.2	30.3	–4.5	–2.7	2.1	3.7
Median		6.0	31.6	31.1	19.6	–15.9	–11.7	–11.2	–17.3
Mean		9.3	27.2	33.5	18.0	–19.4	–12.9	–10.4	–15.5
SD		8.7	17.9	26.3	10.2	13.6	9.3	8.1	10.1

Values represent average enthalpy SD) and temperature of the most pronounced peak for at least five replicates for each species.

Representative DSC scans are provided in Fig. 8.

^aIn table 3 is indicated that the value for O. regalis was obtained from samples stored for 2 yrs instead than 5–12 yrs.

Glass transitions in compositionally complex dried cytoplasm are difficult to detect because the signal is small and can be masked by TAG transitions. To observe glass melting signals throughout the entire moisture range, we used relatively large (6–10 mg) samples of O. regalis spores, which have a small TAG signature. Glass transitions occurred as stepwise functions at a 0.1 mW g^–1 scale (Fig. 9, indicated by asterisks). These events were reliably recognized at temperatures above TAG melting transitions in O. regalis spores conditioned at RH ≤55% and room temperature. Comparable transitions can be observed at temperatures below TAG melting signals for spores placed at RH ≥75%. Water freezing and melting transitions are easily identified by peaks that dominate DSC scans upon minor increases in water content when spores are exposed to RH >90%. Onset of water crystallization of spores moistened at 90% RH occurs at <–45°C (data not shown), and peak temperatures for melting endotherms increase from –25°C to near 0°C as water content increases (Fig. 9).

Fig. 9

Structural transitions in aqueous and TAG domains of Osmunda regalis spores. Water contents are as indicated. All samples were heated at 10°C min^–1 in scanning mode, and the endothermic direction is upwards. Power values from DSC were corrected for sample mass, which was usually between 6 and 10 mg. Shaded area represents temperature range for TAG melting events. * represent glass transitions in aqueous regions of the cell. ← indicates a water melting event.

Lipid content and fatty acid composition

The lipid content of fern spores from tested species was between 0.052 and 0.073 (green spores) and 0.20 and 0.31 (non-green spores) g lipid g^–1 DW Supplementary table S3. Fatty acid composition varied among species and spore types, with green spores containing 2–4 times more saturated fatty acids than non-green spores Supplementary Table S3. Oleic acid was prominent in non-green spores, representing 79% and 52% of the total fatty acids in P. aculeatum and P. vittata, respectively. Linoleic acid content was also high (34%) in P. vittata spores, which were compositionally similar to Thelypteris palustris [data not shown; DSC scans available in Ballesteros and Walters (2007)].

Discussion

This study presents a unique view of time-dependent survival of preserved germplasm. Unlike most labs studying this topic, we focus on conditions well below Tg. We use fern spores from diverse species exhibiting rapid (within a month, Fig. 3A) and slow (over several years, Fig. 1C) aging to probe the effects of moisture and temperature. In particular, we report detectable deterioration of spores stored over LN vapor (approximately –176°C) within a relatively brief period [i.e. ∼4,300 d (∼12 years)] as well as temperature anomalies in deterioration kinetics when spores were stored in conventional freezers (–18°C). The intention is to explain these unexpected results by relating temperature and moisture effects of aging kinetics to current understanding of the coupling between molecular mobility, chemical reactivity and molecular stability within a glassy matrix.

There is a general presumption that LN temperatures are so low that reactions involved in aging cannot occur. Background radiation, deemed to be the largest risk factor for aging of cryopreserved material, had a negligible effect, at least during a short period of study (Glenister and Lyon 1986). However, most data sets are limited to about 20 years (7,300 d) and, combined with difficulty of quantitatively detecting loss of function, we are led to the general perception that time nearly stops during cryopreservation (e.g. Pruksananonda et al. 2012). However, this perception must be balanced with the knowledge that inability to detect change initially is a hallmark of co-operative kinetics that occur in solid materials (Figs. 1–3), and that monitor tests are simply snapshots that do not predict approach to the cliff of rapid mortality that defines the population mean response.

Degradation during LN storage demonstrates that reaction kinetics in solid materials can continue even when global mobility (molecular mobility driving diffusion-based reactions and structural relaxation) is severely limited. In fact, temperature had relatively minor effects on the aging rate in fern spores stored at ≤–18°C (ANCOVA; F = 1.6, df = 1, P = 0.21), which is also evident from an extremely low E_a in Arrhenius plots (E_a <1 kJ mol^–1, Fig. 6; Table 2), and is consistent with previous reports (Walters et al. 2004, Ballesteros et al. 2011). The current experiment avoided possible artifacts from temperature cycling. Moisture also did not have an appreciable influence on aging rate in this temperature range (GLM: df = 3, P = 0.481 for storage temperatures ≤–18°C). Low temperature dependency is also a recognized problem in solid foods and pharmaceuticals, and has been attributed to continuation of local (β) motions (Yoshioka and Aso 2007, Bhattacharya and Suryanarayanan 2009, Chang and Pikal 2009, Laitinen et al. 2013). Interestingly, some chemical additives, such as glycerol, appear to reduce local motion (Chang and Pikal 2009). As an important constituent in cryoprotectant cocktails (Rall and Fahy 1985), glycerol may provide protection to cryopreserved materials beyond its recognized role in supressing ice formation.

Near –20°C, there is a major change in the temperature dependency of fern spore deterioration, and aging rate increases significantly with increases in temperature (ANCOVA, F = 30.2, df = 1, P < 0.001). The change in temperature dependency of deterioration in the solid matrix above –20°C is indicated by a clear break in the Arrhenius plot, marking when the kinetics of aging reactions become dominated by global mobility (i.e. diffusion or α structural relaxation) (Yoshioka and Aso 2007). Values for E_a of spore aging above the Arrhenius break (Table 2) are comparable with diverse dry microbial and plant germplasm reported in this temperature range (Walters et al. 2004, Walters et al. 2005a, Ballesteros et al. 2011). These values are a bit higher than expected for diffusion-controlled reactions in fluids (8–25 kJ mol^–1) and a bit lower than viscosity measured in seeds based on structural relaxation considerations (185 kJ mol^–1 near Tg and 55 kJ mol^–1 near T_K) (Walters 2004).

Glass transitions in dried cytoplasm occur within the –20 to +45°C temperature range (Tg for spores adjusted at 75% and 14% RH, respectively, Fig. 9; see also Ballesteros et al. 2017). Though moisture affects aging rate, it does not affect E_a in this temperature range (e.g Arrhenius slope = 10.7 ± 1.7 and 8.6 ± 0.3 for M. struthiopteris spores treated at 75% and 13% RH, respectively; Table 2). In other words, Arrhenius plots appear parallel (Fig. 5B). We might expect non-linear Arrhenius plots in the vicinity of Tg, according to WLK or VTF models. However, curvilinear fits to aging data offered no higher probabilities or r² than Arrhenius plots (comparisons of fit not shown; results of linear regression are in Table 2), indicating that these mathematically equivalent models compressed to a single slope (Peleg 2017). The linearity of temperature dependency and small heat capacity change (ΔC_p) during glass transitions (Fig. 9) suggests that properties in fern spore cytoplasm exhibit ‘strong,’ rather than ‘fragile’ (sensu Angell 2002) glassy tendencies, meaning that structure above Tg is retained and the extreme increase in viscosity below Tg does not occur. The low heat capacity change during glass transitions measured using DSC (Fig. 9) also supports this conclusion. Linear Arrhenius behavior through Tg provides a powerful tool for using accelerated conditions (i.e. warm temperatures) to predict longevity of materials as long as moisture treatments are similar.

There also appears to be an anomaly at freezer temperatures which is exhibited by some non-green spores (deviation of –18°C points on Arrhenius plots). Spores of P. aculeatum, D. antarctica and P. vittata aged faster during –18°C storage at some moisture levels compared with counterparts stored at 5°C (Figs. 1,6; see P50s in Table 1). Similar observations reported previously led to the general conclusion that freezer storage of fern spores is unreliable (Quintanilla et al. 2002, Li and Shi 2014). In contrast, green spores age fast at all temperatures with no anomaly.

It is tempting to attribute a water transition to the anomaly of freezer-stored materials (Li and Shi 2014). However, low levels of ice formation can only be detected in spores moisturized at RH >90% at room temperature (Fig. 9; Ballesteros and Walters 2007). Instead, there is ample evidence from DSC scans that TAGs crystallize and melt in non-green spores within this temperature range (Figs. 7,8). Green spores, which did not exhibit the anomaly, contained fewer lipids and exhibited smaller TAG transitions than non-green spores (Table 3). TAG crystallization was relatively slow, demonstrated by lower enthalpies for crystallization compared with melting, especially when TAGs nucleate at temperatures <–20°C (cooling scans not provided; Ballesteros and Walters 2007). The increase in melting enthalpy with storage time at –20°C indicates an increased proportion of TAG molecules participating in lipid crystals. Peak shifts to warmer temperatures indicate that the molecular structure of crystals is constantly reorganizing into denser, lower energy forms (Fig. 7; Supplementary Table S3; Himawan et al. 2006). Migration and reorganization of TAG molecules demonstrate that there is considerable diffusive motion within the lipid domain of cytoplasm despite the low temperature and presence of crystals. TAG crystallization is implicated in temperature anomalies reported for intermediate seeds, though the kinetics of damage in fern spores appear to be different (Pritchard and Seaton 1993, Crane et al. 2006, Hamilton et al. 2009). Our working hypothesis is that the continued compression of lipid molecules as they crystallize leads to larger pores in the solid aqueous matrix, increasing the tendency for structural relaxation (Shamblin et al. 1999, Walters 2004, 2015). In other words, crystallization of TAG serves as a plasticizer in the composite matrices of cells.

The temperature dependency of TAG crystallization can be observed by comparing the size of the DSC endotherms for spores stored for the same period but at different temperatures (Fig. 8; Table 3). Transition size for 5°C stored material is, not surprisingly, small because nucleation does not occur at this high temperature. There was more TAG crystal growth in spores stored at –80°C compared with counterparts stored at –20°C, which indicates extensive global mobility of TAGs at –80°C. The considerable TAG crystallization may lead to greater pore size in the aqueous domain of –80°C- compared with –20°C-stored material; notwithstanding, global motion is too restricted for this to have much effect on aging kinetics. Global mobility of the TAG domain at LN temperatures is apparent by the higher enthalpy of melting events in stored, compared with fresh, spores (Fig. 8; Table 3); but TAG mobility is considerably reduced compared with counterparts stored at –20 or –80°C. Collectively, these data illustrate considerable differences in molecular mobility in aqueous and lipid domains of cells. At –20°C, global motion in both domains can lead to anomalously rapid aging and, at LN temperatures, local motion dominates in the aqueous domain, but there is still evidence of global motion in the lipid domain.

Species was also a significant factor of aging rate (ANCOVA, df = 12, P < 0.001). The species effect is exhibited by stacked near-parallel Arrhenius plots (Fig. 5A) having a similar E_a (Table 2). In our analyses, the pre-exponential coefficient (intercept of Arrhenius regression) does not differ significantly among species—as expected for parallel lines—because experimental uncertainty of the slope has a large effect during extrapolation to the y-axis (analysis not shown). The pre-exponential coefficient, explained as the number of ‘diffusional jumps’ needed to effect change (Chang and Pikal 2009), suggests that fewer intermolecular collisions are needed to cause mortality in faster aging cells. Accordingly, faster aging could be attributed to less restricted motion by neighboring molecules, greater proximity of aging substrates or greater vulnerability of critical molecules, all of which are consistent with concepts that implicate both structural (Buitink and Leprince 2008, Ballesteros and Walters 2011) and compositional (e.g. Buitink and Leprince 2008, Birtić et al. 2011, Sano et al. 2016, Pereira Lima et al. 2017) factors. The single-cell nature of fern spores and their tendency to deteriorate in an experimentally tractable time frame makes them ideal for evaluating features that explain tendency to degrade rapidly.

Summary

Preserving viability of biological materials necessarily requires their transformation from fluid to solid. Deterioration continues in a preserved system, albeit at slow rates that are regulated by molecular mobility. Fern spores provide an experimentally tractable system to explore coupling between molecular mobility and deterioration kinetics. We show moderate temperature dependency for aging reactions through the temperature range where glass transitions occur, suggesting that biological solids resemble ‘strong’ glasses. However, temperature dependency decreases by two orders of magnitude below freezer conditions, indicating that the global mobility of diffusive motion has little role in the aging rate. Faster than expected aging during freezer storage was occasionally observed and is attributed to plasticization of aqueous glasses when TAGs, which have different temperature dependencies compared with cytoplasm, crystallize. Diverse spore species showed a similar pattern of temperature dependency despite extreme differences in aging rate. Longevity may depend on protective mechanisms that separate substrates in aging reactions or reduce the probability of a damaging intermolecular interaction.

Materials and Methods

Plant materials

Fern spores from eight species were collected from 2005–2016 from wild populations in Spain, the USA and Portugal or from plants cultivated in botanic gardens Supplementary Table S2. Mature fronds, with mature spores still enclosed in the sporangia, were removed from plants and immediately placed on poster paper and dried for 1–5 d under ambient conditions in the laboratory. After sporangial dehiscence, spores were collected from the paper, sieved and subsequently stored at 4°C until preparation for storage or experimentation, which was usually within 1 week (green spores, except M. struthiopteris that was stored at 4°C for 26 d) and 6 months (non-green spores) after spore collection Supplementary Table S2. Spores from Spain and Portugal were packed in an insulated box and mailed via expedited post to NLGRP in Fort Collins, CO, USA, where storage experiments occurred.

Moisture and temperature controls during storage

The experiment included both moisture and temperature variables in a multifactorial design. Moisture was manipulated at room temperature by placing samples over several saturated salt solutions that controlled RH from 1% (P₂O₅) to 85% (KCl), with the most common treatments including 1% (P₂O₅), 14% (LiCl), 33% (MgCl₂), 55% [Ca(NO₃)₂] and 75% (NaCl) RH (Ballesteros and Walters 2007). Water adjustment occurred over a 10 or 5 d period in spores treated before and after 2009, respectively (Ballesteros et al. 2017). RH was monitored and confirmed using Hobo dataloggers, model # U12-011 (Onset Corporation). About 0.1–5 mg aliquots of moisture-adjusted spores were then hermetically sealed into 25 µl volatile aluminum DSC pans (Perkin Elmer) that could be retrieved individually for subsequent testing (Ballesteros and Walters 2007). Pans were then placed in freezers or vapor above LN to achieve storage temperatures of –18, –80 and about –176°C. Before 2009, the same moisture adjustment and packaging procedures were used for spores stored in incubators set at 45, 35, 25 and 5°C. After 2009, RH, rather than water content, was controlled, and this was done by maintaining RH chambers, with spores held in Petri plates or open vials, in the incubators (Ballesteros et al. 2017). Storage in vapor above LN (i.e. ‘cryogenic storage’) used an MVE ‘stock series’ cryotank (1,400 liter LN capacity) (Chart MVE Biomedical).

Viability assessments from germination in vitro

Ability to germinate or effect normal development is the ‘gold standard’ technique for assessing health and viability of preserved plant propagules including fern spores (e.g. Quintanilla et al. 2002, Ballesteros et al. 2011, Li and Shi 2014, Ballesteros et al. 2017). In vitro germination techniques using appropriate culture medium and germination conditions were previously reported for fern spores and there was no indication of specialized germination requirements in species used for this study (e.g. Quintanilla et al. 2002, Ballesteros et al. 2011, Ballesteros et al. 2017). The feasibility of using tetrazolium tests (ISTA 2018) to speed germination assays or confirm death was tested in some of the species (Supplementary Table S1), but in vitro germination assays were deemed more broadly applicable to diverse species.

Survival of spores during storage was assayed periodically to give between three and eight time points used to assess aging kinetics. Three time points were the minimum required to assess longevity with the two models employed (see below). The sampling period was adjusted according to the expected aging rate, being more frequent when spores were expected to age faster. For example, viability testing within a species occurred more often (every 3–5 d) in spores stored at 45°C and was more spaced [every 360–1,000 d (1–3 years)] in spores stored below 0°C. Assays giving unexpectedly high or low viability assessments were repeated within a few weeks for confirmation, and results of these narrow time intervals were combined. Samples stored below 0°C or above ambient were brought to room temperature by placing vials or DSC pans on the benchtop for 30 min to 1 h. Spores were then sown into three or four 60 mm diameter Petri dishes filled to half depth with culture medium solidified with agar and prepared with the fungicide nystatin (100 U ml⁻¹) (Ballesteros et al. 2017). Petri dishes were sealed with Parafilm (American National Can) and placed in a growth chamber set at 20 ± 2°C with 12 h light/dark cycles (Ballesteros et al. 2017).

Germination was scored when the outer wall of the spore burst and the rhizoid or the first chlorophyllic cell emerged (Ballesteros et al. 2017). This was observed using a dissecting microscope at ×40 magnification. Over 100 spores were located and scored per Petri dish, and the number of germinated spores per 100 observed per dish was recorded throughout the germination period. Spores were selected randomly in each Petri dish, often using several fields of view within each dish. Viability was usually assessed near days 7 and 15 for green spores and days 20 and 30 for non-green spores, and the higher of the two values was used as the viability for that storage treatment and Petri dish (Ballesteros et al. 2017). After verifying no effect among Petri dishes in a treatment (usually arising from fungal contamination), data among Petri plates within a treatment were pooled, to calculate the proportion of germinating spores (Crawley 2007) for each storage treatment and duration. This value is termed total germination and is depicted as N_t in kinetic models. At higher storage temperatures, experiments were usually terminated when germination declined to 0%, though there are some incidences when the sample was depleted before complete deterioration was observed. For storage at –18°C and below, samples are retained in storage at NLGRP and are available for conservation goals or analyses after longer storage times.

Calculations of longevity and temperature dependency of survival

Longevity of spores of different species stored under different conditions was calculated from changes in germination (N_t) with storage time (t). Longevity was expressed as P50 (time for initial germination to decline to half that value), a parameter that accounts for both the duration of the asymptomatic phase and the rapidity of viability loss once that phase commences (Walters et al. 2004). Aging rate was expressed as the reciprocal of P50 (P50^–1). Two models were used to calculate P50, both allowing flexibility in time course shape from exponential, sigmoidal or linear. For the Avrami function [ln(N_t/N₀) = n·(t/ϕ), where N₀ = initial or maximum germination for the species], a linear regression of ln(t) and ln[ln(N₀/N_t)] was used to calculate Avrami coefficients: n (slope of linear regression), ϕ [ϕ = exp(–y₀/n) and y₀ = the y-intercept of linear regression] (Walters et al. 2004, Ballesteros et al. 2017). Values for N₀ for each species were calculated from the average of the five highest viability measurements across all treatments for the species. Values for N_t/N₀ were constrained between 0.995 and 0.001 by constraining values for ln[ln(N₀/N_t)] between –5.3 and 1.93 during curve fitting (e.g. Ballesteros et al. 2017). P50 could be calculated by interpolation when germination declined below 50% of original [P50 < 3,000–4,000 d (∼8–11 years)]. However, we needed to extrapolate models when insufficient deterioration occurred, and this required a ≥10% decline in germination to obtain negative slopes in fitted Avrami models (Walters et al. 2004). Because extrapolation has intrinsic uncertainties, we also fit data to logistic functions using binomial error distributions and used dose response functions to calculate P50 (e.g. Ballesteros et al. 2017).

The effects of temperature on longevity were characterized using classic Arrhenius plots where rate = A₀·exp^(–Ea/RT), with T = temperature (in Kelvin), R = 8.314 J K^–1 mol^–1 (the ideal gas constant), A₀ being a pre-exponential factor and E_a considered the apparent activation energy or temperature coefficient (Buitink et al. 1999, Walters et al. 2004, Ballesteros et al. 2011). Seemingly linear relationships of ln(P50^–1) and T^–1 were used to estimate E_a in order to make inferences about the temperature dependency of the aging rate. Linear relationships were calculated using Excel and an F distribution model with probability set at 0.05.

Differential scanning calorimetry

The nature of structural changes in fern spores during storage was probed between +50 and –150°C using DSC. Analyses were performed on samples soon after harvest (2005–2007, 2009–2010 and 2013) and then in 2017–2018. Freshly harvested samples were prepared as described above: about 0.1–5 mg of samples exposed to RH between 1% and 75% RH at room temperature were hermetically sealed into 25 µl aluminum pans (Perkin Elmer) and stored in 1.5 ml polypropylene cryovials at 5°C (2005–2007 harvested spores only), –18, –80°C and vapour phase above LN (about –176°C). A walk-in refrigerated room held at 5°C was used in 2017–2018 to seal hermetically spores that had been stored in RH-controlled chambers at 5°C since 2010. In preparation for DSC measurements in 2017–2018, pans were removed from storage and immediately plunged into LN. They were then loaded onto the DSC sample holder that had been pre-cooled to –170°C. This procedure was intended to prevent warming of the sample, and estimates from the DSC temperature controller suggested that samples did not warm above –120 to –140°C when the procedure was applied correctly. Based on prior measurements using thermocouples imbedded into punctured pans, we estimate that cooling rates achieved by plunging pans into LN is of the order of 300°C min^–1 (data not shown) between room and –100°C, and 80–150°C min^–1 until the sample reaches –170°C.

Energy associated with transitions was recorded during warming at 10°C min^–1 using one of three DSC7s or a DSC8000 (Perkin Elmer), all calibrated for temperature and energy using appropriate standards for low temperature operation. LN was used as the coolant, and helium was used as the purge gas. After the cold-load run, pans were recooled at 10°C min^–1 to –150°C and rewarmed in scanning mode (the treatment usually reported in the literature), and then cooled again to –150°C after a 60 min annealing treatment at –20°C. Following this procedure, pans were placed in a conventional freezer (–20°C) and stored for 4–15 d before they were cold-loaded into the calorimeter and scanned again. In 2017–2018, between four and eight DSC scans were collected per species per storage temperature, except for EQ, MS and CM (2010 harvest from Galicia) samples stored at 5°C, for which there were few or no samples remaining. Enthalpy was calculated as the area of a peak using Perkin Elmer software. DSC scans of recooled + rewarmed samples that had been stored at –80 and –176°C were similar, allowing us to pool enthalpy values among these storage treatments to calculate average characteristics for samples cooled at 10°C min^–1 or annealed at –20°C for 1 h to several days. Enthalpy calculations were performed using Perkin-Elmer Pyris software. Enthalpy is expressed on a per g DW basis, unless indicated otherwise.

Analysis of lipid content and fatty acid composition by GC

Analysis of lipid composition followed standard procedures of Bligh and Dyer lipid extraction and GC characterization of fatty acid methylesteraases (FAMEs). Total lipids were extracted from a pre-weighed sample of approximately 50 mg of spores, that had been stored at –18°C, using repeated washes of chloroform:methanol (2:1) solution. Lipids separated into the chloroform phase and recovered by evaporating the solvents using a stream of N₂ gas. Fatty acids from extracted lipids were derivatized using 14% BF₃ (Sigma) in methanol heated to 92–95°C for 3 min. Water (1 ml) and petroleum ether (2 ml) were then added to the cooling solution and the organic phase (top layer) was pipetted off, dried under N₂ and resuspended in 0.5 ml of chloroform. Between 0.25 and 0.5 ml of the chloroform:FAME solution was separated by GC (Perkin Elmer 8500) using a Nukol fused silica capillary column (30 m, 0.25 mm i.d.; Supelco). Oven temperature was set at 100°C and ramped at 10°C min^–1 to 190°C. Injector and detector (FID) temperatures were set at 220°C. FAMEs were identified by retention time using FAME RM3 standards (Supelco). Peak height was used to determine proportions of each FAME according to instrumentation software.

Statistical comparisons

All survival data were treated as proportion data according to Crawley (2007). Significant differences between germination before (also indicated as initial germination) and after moisture equilibration were analyzed using a two-way binomial test using GenStat^® 14th edition (VSN International Ltd. 2011). Genstat software 14.2 (VSN International Ltd. 2011) was also used to fit germination/storage time data to logistic functions using binomial error distributions and to calculate P50. This approach enabled us to provide error estimates and statistically compare effects of species, temperature and moisture on longevity. Significance of storage time, storage temperature and moisture treatments on spore viability was tested using GLMs and a binomial error distribution available in Genstat software 14.2 (VSN International Ltd. 2011) or IBM SPSS Statistics 19 (IBM Corp. 2010). Correspondence between P50 values calculated by logistic functions or Avrami models (Table 1) were tested by correlation using the Pearson coefficient (r).

Funding

This work was supported by the Spanish Ministry of Science and Technology [a Formación del Profesorado Universitario (FPU) grant (to D.B. 2006)] and the Agricultural Research Service (ARS) [an ARS-funded post-doctoral fellowship (2009–2012) and appropriated funding to the ARS (2010–current)]. Royal Botanic Gardens Kew receives grant-in-aid from Defra, UK.

Acknowledgments

The authors thank J.E. Larson, C. Mayer, E. Estrelles, A.M. Ibars, R. Botelho, L.G. Quintanilla and V.C. Pence for their help and support collecting fern spores used in the experiments. USDA is an equal opportunity provider and employer. Mention of trade names or commercial products in this article is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the US Department of Agriculture.

Disclosures

The authors have no conflicts of interest to declare.

Footnotes

Subject areas: (2) environmental and stress responses