Comparative metabolic profiling between desiccation-sensitive and desiccation-tolerant species of Selaginella reveals insights into the resurrection trait




Spike mosses (Selaginellaceae) represent an ancient lineage of vascular plants in which some species have evolved desiccation tolerance (DT). A sister-group contrast to reveal the metabolic basis of DT was conducted between a desiccation-tolerant species, Selaginella lepidophylla, and a desiccation-sensitive species, Selaginella moellendorffii, at 100% relative water content (RWC) and 50% RWC using non-biased, global metabolomics profiling technology, based on GC/MS and UHLC/MS/MS2 platforms. A total of 301 metabolites, including 170 named (56.5%) and 131 (43.5%) unnamed compounds, were characterized across both species. S.  lepidophylla retained significantly higher abundances of sucrose, mono- and polysaccharides, and sugar alcohols than did S. moellendorffii. Aromatic amino acids, the well-known osmoprotectant betaine and flavonoids were also more abundant in S. lepidophylla. Notably, levels of γ-glutamyl amino acid, linked with glutathione metabolism in the detoxification of reactive oxygen species, and with possible nitrogen remobilization following rehydration, were markedly higher in S. lepidophylla. Markers for lipoxygenase activity were also greater in S. lepidophylla, especially at 50% RWC. S. moellendorffii contained more than twice the number of unnamed compounds, with only a slightly greater abundance than in S. lepidophylla. In contrast, S. lepidophylla contained 14 unnamed compounds of fivefold or greater abundance than in S. moellendorffii, suggesting that these compounds might play critical roles in DT. Overall, S. lepidophylla appears poised to tolerate desiccation in a constitutive manner using a wide range of metabolites with some inducible components, whereas S. moellendorffii mounts only limited metabolic responses to dehydration stress.


The Selaginellaceae (spike mosses) represent one of two major vascular plant lineages, the lycophytes and the euphyllophytes, that arose during the Silurian. The evolution of lycophytes was accompanied by the acquisition of roots with root caps, and typically simple and single-veined leaves called microphylls (Friedman, 2011). Lycophytes dominated the world’s flora from the Devonian through the Carboniferous to the end of the Permian, which corresponded with a dramatic decline in atmospheric CO2 abundances, and now provide an important energy source in the form of coal (Beerling, 2002; Gensel, 2008). Just over 1000 species of extant lycophytes form a sister clade, monophyletic to other vascular plants, that includes the Lycopodiaceae (club mosses), the Isoeteaceae (quillworts) and the Selaginellaceae (spike mosses), found within a single Selaginella genus containing approximately 700 species (Banks, 2009). Over the last 400 million years, members of this family have come to occupy diverse habitats, including arctic, temperate, tropical and semi-arid environments (Banks, 2009). Molecular phylogenetic analysis within this genus based on nuclear and plastid DNA sequence data has revealed remarkably high rates of DNA sequence divergence relative to angiosperm families (Korall and Kenrick, 2004; Little et al., 2007). Selaginella moellendorffii has one of smallest plant genomes known to date (∼110 Mb; Wang et al., 2005), and has recently been sequenced (Banks et al., 2011). Expressed sequence tag collections have been reported for several Selaginella species, including S. moellendorffii (Weng et al., 2005) and Selaginella lepidophylla (Iturriaga et al., 2006).

Several members of the Selaginella genus, including S. lepidophylla (Iturriaga et al., 2006), Selaginella bryopteris (Deeba et al., 2009) and Selaginella tamariscina (Wang et al., 2010), have evolved desiccation tolerance (DT) or the ability to survive air drying (Oliver et al., 2000a), whereas most lack the resurrection trait. Although this trait is relatively common in algae, lichens and mosses (Alpert, 2006; Wood, 2007), it is extremely rare among vascular plants (∼0.15% of species exhibit DT; Oliver et al., 2000a; Porembski and Barthlott, 2000; Proctor and Pence, 2002; Proctor and Tuba, 2002).

Integrative functional genomic approaches have been employed recently to elucidate the mechanistic basis of DT (Moore et al., 2009; Cushman and Oliver, 2011; Farrant and Moore, 2011). In contrast to gene discovery and proteomic studies, relatively few ‘omics’-scale studies of metabolites have been undertaken in resurrection plants, although the abundance of several key metabolites have been reported. For example, sucrose is the most constitutively abundant sugar in desiccation-tolerant bryophytes (Smirnoff, 1992) and ferns (Farrant et al., 2009), and accumulates during drying in desiccation-tolerant angiosperms (Ghasempour et al., 1998; Cooper and Farrant, 2002; Whittaker et al., 2007; Phillips et al., 2008; Toldi et al., 2009; Oliver et al., 2011). Sucrose is the second most abundant sugar in desiccation-tolerant Selaginella species, which accumulate both sucrose and trehalose in desiccated and hydrated tissues (Adams et al., 1990; Iturriaga et al., 2000; Vázquez-Ortíz and Valenzuela-Soto, 2004; Liu et al., 2008). In contrast, glucose and fructose accumulate in resurrection plants, but their abundance typically decreases as dehydration progresses in bryophytes (Smirnoff, 1992), lycophytes (Vázquez-Ortíz and Valenzuela-Soto, 2004), ferns (Farrant et al., 2009) and angiosperms (Ghasempour et al., 1998; Whittaker et al., 2001, 2007; Oliver et al., 2011). The reduced abundance of these two sugars during the dehydration process might result from phosphorylation and metabolic processes, or from conversion into protective sugars (Whittaker et al., 2001, 2007; Farrant et al., 2009). Raffinose-series sugars also accumulate in ferns (Farrant et al., 2009) and angiosperms capable of DT, where they might play roles in stress protection and carbon storage (Ghasempour et al., 1998; Norwood et al., 2000; Živkovićet al., 2005; Peters et al., 2007; Oliver et al., 2011).

Many amino acids, particularly those rich in nitrogen, such as asparagine, arginine, glutamate and glutamine, accumulate during dehydration in Sporobolus stapfianus, where they may function as compatible solutes or as mobile nitrogen sources (Martinelli et al., 2007; Whittaker et al., 2007; Oliver et al., 2011). Changes in membrane lipid content and composition that might alter membrane fluidity and reduce solute leakage through the lipid bilayer during desiccation have been observed in some dessication-tolerant species (Quartacci et al., 2002; Oliver et al., 2011).

To provide a comprehensive understanding of the role of metabolites that form the basis of DT in a key vascular plant lineage that diverged over 400 million years ago, the metabolomes of two closely related species of Selaginella that differ in their sensitivity to desiccation were compared at 100 and 50% relative water content (RWC) using an unbiased, high-throughput metabolomics platform. This sister-group comparison revealed that S. lepidophylla is poised metabolically to tolerate desiccation in both a constitutive and an inducible manner, whereas S. moellendorffii fails to adapt metabolically to dehydration stress and dies.

Results and Discussion

In-depth knowledge of metabolites and their production and degradation provide a detailed and highly informative phenotype of an organism (Wiechert et al., 2007; Cascante and Marin, 2008). Although large-scale metabolomics studies have been reported for Arabidopsis plants undergoing dehydration (Urano et al., 2009), similar studies have not yet been performed for any resurrection species, except for a recent comparative analysis of two closely related Sporobolus species that differed in DT (Oliver et al., 2011). In the current study, a total of 301 metabolites were identified to provide insights into the specific metabolic pathways essential for the acquisition of DT among early land plants (i.e. lycophytes) that diverged from euphyllophytes more than 400 million years ago (Friedman, 2011).

Sister Selaginella species respond differently to dehydration

Although S. lepidophylla and S. moellendorffii share similar anatomical features, they have evolved very divergent responses to dehydration. The dehydration curves of the two Selaginella species differed from one another, with S. moellendorffii (dessication sensitive, DS) exhibiting a more rapid rate of water loss than S. lepidophylla (DT; Figure 1a). S. moellendorffii reached 25% RWC within 1 h, whereas S. lepidophylla required 4 h to reach this same point. This rate is similar to S. bryopteris, a related resurrection species that takes 6 h to lose 80–90% of its RWC (Deeba et al., 2009; Pandey et al., 2010). Stem curling, an adaptive response to protect the plant from damage caused by high irradiance, high temperatures or both, under either laboratory (Lebkuecher and Eickmeier, 1991, 1992) or field conditions (Lebkuecher and Eickmeier, 1993), was evident at 50% RWC in both species (Figure 1b). After 24 h of drying, S. lepidophylla retained 7% RWC, whereas S. moellendorffii retained only 1.5% RWC. In the dried state S. lepidophylla retained apparent chlorophyll, indicating that it is homiochlorophyllous.

Figure 1.

 Dehydration rates in Selaginella lepidophylla and Selaginella moellendorffii.
(a) Dehydration time courses for S. moellendorffii (blue line and diamonds) and S. lepidophylla (red line and boxes), measured by tracking the percentage loss in relative water content (RWC) over time (0, 1, 2, 3, 4, 6, 12 and 24 h). Data represent the mean of three independent replicate experiments (n = 9). Error bars represent standard error of the mean.
(b) Representative images of S. lepidophylla (SL) and S. moellendorffii (SM) plants collected at 100% RWC (100) and 50% RWC (50). Scale bars: 2 cm.

The two species also displayed differential responses to rehydration. S. lepidophylla can survive years of desiccation, whereas S. moellendorffii cannot recover fully from water losses at, or below, 40% RWC (Figure S1). Furthermore, as with drying, S. moellendorffii displayed a very rapid rate of rehydration, with full rehydration occurring within 2 h. In contrast, S. lepidophylla rehydration required 24 h to complete the rehydration process (Figure S2).

Metabolome composition in S. moellendorffii and S. lepidophylla

To compare the metabolomic response to dehydration between S. moellendorffii and S. lepidophylla, six biological replicates from plants sampled at full hydration (100% RWC) and partial dehydration (50% RWC) were analyzed from each species using non-biased, global metabolome technology based on GC/MS and UHLC/MS/MS2 platforms (Evans et al., 2009). The combined platforms detected a total of 301 metabolites, of which 170 (56.5%) were named and 131 (43.5%) were unnamed (Figure 2; Table S1). Slightly more named compounds could be identified in S. lepidophylla (165 or 54.8%) than in S. moellendorffii (148 or 49.1%). Upon mapping the named compounds to biochemical pathways, the percentages within categories varied little between the two sister species, with amino acids having the highest percentage of individual components (15–16%), followed closely by carbohydrates (15%), lipids (11–13%), co-factors (6%), nucleotides (5%), peptides (2–4%) and secondary metabolites (2%; Figure 2). Slightly more unnamed compounds were identified in S. moellendorffii (117 or 44%) than in S. lepidophylla (102 or 38%).

Figure 2.

 Functional categorization of named and unnamed metabolites detected at either 100% or 50% RWC: (a) 265 metabolites from Selaginella moellendorffii and (b) 267 metabolites from Selaginella lepidophylla.

Metabolomic differences between S. moellendorffii and S. lepidophylla

Partial least squares-discriminant analysis (PLS-DA) was used to build a comparative model of the 100 and 50% RWC hydration states between S. moellendorffii and S. lepidophylla. Metabolites missing three or more of six replicate (50%) values in either species were excluded, resulting in a total of 199 metabolites for comparison. The first three PLS-DA components explained 63.9% of the variation (explained variation, R2 = 0.75; predicted variation, Q2 = 0.47), and showed distinct clustering between the two species at both 100 and 50% RWC, suggesting that various metabolites were likely to account for the differences observed in the model (Figure 3). Interestingly, S. lepidophylla samples showed a much smaller separation between hydration states than S. moellendorffii, indicating a metabolic predisposition at 100% RWC for responses associated with water deficit stress adaptation at 50% RWC. In this regard, S. lepidophylla appears to resemble desiccation-tolerant byrophytes, which rely upon constitutive cellular protection coupled with a rehydration-induced recovery process that involves cellular repair (Oliver et al., 2005). However, the S. lepidophylla models at each RWC are also distinct from one another, indicating that vegetative DT might also involve some environmental induction of cellular protection processes, as employed by the angiosperms (Moore et al., 2009). The use of both constitutive and inducible DT strategies is consistent with the intermediate phylogenetic position of lycophytes between bryophytes and angiosperms.

Figure 3.

 Metabolite distribution in Selaginella lepidophylla and Selaginella moellendorffii, as defined by partial least-squares discriminant analysis (PLS-DA) constructed from 199 metabolites.
The plots show the three main components that explain 63.9% of the variation (R2 = 0.75; Q2 = 0.47). Squares (blue) and crosses (pale blue) indicate 100 and 50% RWC in S. moellendorffii, respectively; asterisks (green) and upside down triangles (red) indicate 100 and 50% RWC in S. lepidophylla, respectively.

In comparing the metabolite profiles between S. moellendorffii and S. lepidophylla at 100% RWC, a total of 190 compounds were detected that showed significantly altered abundances (P < 0.05), including 117 (52 named and 65 unnamed) that were more abundant in S. moellendorffii and 73 (48 named and 25 unnamed) that showed a greater abundance in S. lepidophylla. Amino acids and unnamed compounds were clearly over-represented in S. moellendorffii. In contrast, various carbohydrates, particularly 4C, 5C and 6C sugars, and sugar alcohols, were over-represented in S. lepidophylla in the fully hydrated state (Table S1). At 50% dehydration, a total of 197 compounds with significantly altered abundances (P < 0.05) were detected, including 112 (52 named and 60 unnamed) that were more abundant in S. moellendorffii and 85 (54 named and 31 unnamed) that showed a greater abundance in S. lepidophylla. Under partial dehydration conditions, γ-glutamyl amino acids were clearly over-represented in S. lepidophylla, whereas unnamed compounds were clearly over-represented in S. moellendorffii (Table S1).

Energy metabolism

Obvious differences were observed in the abundances of metabolites of glycolysis/gluconeogenesis, the tricarboxylic acid (TCA) cycle and sugar metabolism. The relative levels of lactate, which is derived from pyruvate, were significantly (P < 0.05) greater in S. lepidophylla (Figure 4). Two glycolysis/gluconeogenesis intermediates, glucose-6-phosphate (G6P) and fructose-6-phosphate (F6P), increased during dehydration in S. moellendorffii to significantly (P < 0.05) exceed the levels present in S. lepidophylla at 50% RWC. Other compounds, such as ribose, galactose and 3-deoxyoctulosonate, were also significantly (P < 0.05) greater in S. moellendorffii for both hydration states (Table S1). Many TCA cycle intermediates (e.g. citrate, fumarate and malate) were also significantly (P < 0.05) more abundant in S. moellendorffii in both hydration states. However, two (e.g. cis-aconitate and succinate) were significantly (P < 0.05) more abundant in S. lepidophylla, but only at 100% RWC, whereas isocitrate was the only TCA cycle intermediate with relative levels significantly (P < 0.05) greater in S. lepidophylla at both 100 and 50% RWC.

Figure 4.

 Differences in the metabolites involved in energy metabolism between Selaginella lepidophylla (SL) and Selaginella moellendorffii (SM).
Compounds in red or green indicate greater relative abundances in S. lepidophylla or S. moellendorffii, respectively. Dotted arrows indicate a gap in the pathway and double-headed arrows indicate a reversible reaction. The x-axis represents the percentage of relative water content (% RWC) at 100% RWC (100) and 50% RWC (50). The y-axis box plots indicate the scaled intensity mean (+) and median (—) values: top/bottom ranges of boxes indicate upper/lower quartiles, respectively; top/bottom whiskers indicate the maximum/minimum distribution of the data. The P and Q values for the comparisons are presented in Table S1. G6P, glucose-6-phosphate; F6P, fructose-6-phosphate.

Carbohydrate metabolism

Carbohydrates are among the best-studied metabolites in resurrection plants, and accumulate in vegetative tissues in response to drying in all desiccation-tolerant tracheophytes studied (Alpert and Oliver, 2002). Sucrose, in combination with other sugars, such as raffinose-series sugars, are thought to protect vegetative tissues from damage by forming intracellular glasses and by replacing water molecules to prevent membrane fusion during drying (Koster, 1991; Hoekstra et al., 2001; Sakurai et al., 2008). For example, sucrose accumulates in all desiccation-tolerant fern (Farrant et al., 2009) and angiosperm species examined to date (Ghasempour et al., 1998; Peters et al., 2007; Toldi et al., 2009). In Sporobolus stapfianus, sucrose accumulated steadily throughout drying with 10.5-fold higher abundances in dried than in hydrated plants (Oliver et al., 2011). In contrast, desiccation-tolerant mosses maintain constitutively high levels of sucrose that do not increase during drying (Smirnoff, 1992).

Glucose, sucrose and 1-kestose, a fructotrisaccharide, were significantly (P < 0.05) more abundant at 2.0-fold, 5.6-fold and 12.5-fold, respectively, in S. lepidophylla than in S. moellendorffii at 100% RWC (Figure 4; Table S1). These sugars were also more abundant at 50% RWC. However, these differences were not significant, as these sugars increased in abundance in S. moellendorffii, and remained unchanged in abundance in S. lepidophylla during dehydration (Figure 4). Thus, S. lepidophylla appears to employ the same constitutive protection strategy as desiccation-tolerant mosses.

Interestingly, trehalose, a non-reducing disaccharide, was significantly (P < 0.05) more abundant in S. moellendorffii for both hydration states (Figure 4), but declined slightly upon dehydration in S. lepidophylla, consistent with previous observations (Adams et al., 1990). Unlike most other vascular desiccation-tolerant species, lycophytes (e.g. S. lepidophylla, Selaginella sartorii and S. tamariscina) accumulate large quantities of trehalose in both hydrated and dehydrated states (Adams et al., 1990; Iturriaga et al., 2000; Liu et al., 2008), a feature they share with eubacteria, fungi (i.e. yeast) and invertebrates (Elbein et al., 2003). The large accumulation of this non-reducing sugar in S. moellendorffii indicates that trehalose is not necessarily required for DT in lycophytes. Similar results have been found in lower-order organisms. For example, in Saccharomyces cerevisiae, although dehydration correlates well with increased trehalose abundance, mutants that fail to produce trehalose were able to recover from desiccation (Ratnakumar and Tunnacliffe, 2006).

Polyol metabolism

Perhaps the most striking difference in carbohydrate metabolism between the two species was in sugar alcohol metabolism in the fully hydrated state (Figure 5). Most notably, sorbitol abundance was 367- and 146-fold greater at 100 and 50% RWC, respectively, in S. lepidophylla. Similarly, xylitol, arabitol and erythritol, were 117-, 40- and 12-fold greater, respectively, at 100% RWC. These polyols were also greater in S. lepidophylla at 50% RWC. Glycerol abundances were also greater in S. lepidophylla at both 100 and 50% RWC, but these differences were not significant (Table S1). Gluconate and arabonate abundances were 49- and 6-fold greater, respectively, in S. lepidophylla at 100% RWC, and were also significantly (P < 0.05) greater at 50% RWC. Inositol 1-phosphate and myo-inositol were more abundant in S. lepidophylla at both hydration states, but the difference was only significant (P < 0.05) for myo-inositol. The abundance of most sugar alcohols decreased in S. lepidophylla at 50% RWC, with the exception of myo-inositol and inositol 1-phosphate (Figure 5). Mannitol was significantly (P < 0.05) more abundant in S. lepidophylla at 100% RWC (6.8-fold), whereas mannitol increased significantly (P < 0.05) by 3.5-fold in S. moellendorffii at 50% RWC. In contrast, the relative levels of other compounds, such as galactose, 3-deoxyoctulosonate and ribose were significantly (P < 0.05) greater in S. moellendorffii for both hydration states (Table S1).

Figure 5.

 Differences in sugar alcohols and other sugars between Selaginella lepidophylla (SL) and Selaginella moellendorffii (SM).
Compounds in red indicate greater relative abundance in S. lepidophylla. Dotted arrows indicate a gap in the pathway and double-headed arrows indicate a reversible reaction. Box plots are as described in Figure 4. The P and Q values for the comparisons are presented in Table S1. E4P, erythrose-4-phosphate; F6P, fructose-6-phosphate; G6P, glucose-6-phosphate; Ru5P, ribulose-5-phosphate; T4P, threose-4-phosphate; Xu5P, xylulose-5-phosphate.

This massive accumulation of these polyols in S. lepidophylla compared with S. moellendorffii in both hydration states is consistent with several proposed roles for polyols. Firstly, several polyols have been shown to act as osmoprotectants by stabilizing protein structure against (thermal) denaturation, with sorbitol and xylitol showing the best ability to stabilize yeast hexokinase A (Tiwari and Bhat, 2006). This stabilization is believed to occur via preferential hydration of proteins in their denatured state (Xie and Timasheff, 1997), brought about by an increase in surface tension in the presence of polyols (Kaushik and Bhat, 1998). Remarkably, sorbitol and xylitol were the two polyols that exhibited the greatest relative abundance in S. lepidophylla compared with S. moellendorffii, consistent with their potential role in protein stabilization during dehydration and/or desiccation. Secondly, polyols, especially glycerol, display strong water-binding activity via hydrogen bonding (Kataoka et al., 2011), and reduce the surface tension of water (Tiwari and Bhat, 2006), suggesting that these polyols might slow the rate of water loss during the drying process, as observed here (Figure 1). The combined effects of preferential hydration of (denatured) proteins and reduced rates of water loss would be critical during the desiccation process. Slowing the rate of water loss might also allow more time for the biosynthesis of desiccation-adaptive proteins and metabolites, which is known to be required for the establishment of DT (Alpert and Oliver, 2002). Additionally, these polyols might also slow the rate of water absorption following rehydration: notably S. lepidophylla regained 80% RWC within 4 h of rehydration, consistent with this hypothesis. In contrast, S. mollendorfii regained 80% RWC within only 15 min (Figure S2). Several polyols, including arabitol, erythritol and mannitol, were several-fold more abundant in the fully hydrated leaves of desiccation-tolerant Sporobolus stapfianus compared with desiccation-sensitive Sporobolus pyramidalis, indicating that the constitutive expression of selected polyols might be a widespread strategy adopted by resurrection species (Oliver et al., 2011). Thirdly, some sugar alcohols, such as myo-inositol, mannitol and sorbitol, actively scavenge hydroxyl radicals and might help protect against oxidative damage (Sanchez et al., 2008). Lastly, sugar alcohols might participate in stress signaling by stimulating the expression of stress-adaptive genes. For example, Arabidopsis plants engineered to express mannitol were pre-conditioned to abiotic stress through the expression of a variety of stress-inducible genes, including those associated with abscisic acid responses, redox regulation and cell wall strengthening (Chan et al., 2011).

Although myo-inositol and inositol-1-phosphate abundances were higher in S. lepidophylla than in S. moellendorffii (Figure 5), raffinose-series sugars, such as raffinose and stachyose, were not detected. This is in contrast to the constitutive or drying-induced accumulation of galactinol and raffinose in the desiccation-tolerant fern Mohria cafforum (Farrant et al., 2009) and Sporobolus stapfianus, in which these sugars were detected, and had accumulated to over 60-fold increased levels, late in the drying process (>60% RWC; Oliver et al., 2011). Thus, the accumulation of raffinose-series sugars does not appear to be essential for DT in S. lepidophylla. The lack of detectable raffinose-series sugars might appear somewhat surprising, but perhaps the functional role(s) of these sugars has been replaced by other sugars or other compounds with reactive oxygen species (ROS)-scavenging activity, as these sugars are known to act as ROS scavengers in Arabidopsis (Nishizawa et al., 2008; Nishizawa-Yokoi et al., 2008). The absence of raffinose-series sugars might be a peculiar trait of this ancient lineage of plants; however, other desiccation-tolerant lycophytes should be investigated to determine whether or not this is a true evolutionary difference across this key phylogenetic group.

Amino acids, amino acid derivatives and peptides

Of the 46 amino acids and derivatives detected in this study, the relative levels of only five (i.e. betaine, 3-methoxytyrosine, tyrosine, alanine and leucine) were significantly (P < 0.05) greater in S. lepidophylla at 100% RWC, whereas 19 amino acids were significantly (P < 0.05) more abundant in S. moellendorffii (Table S1). At 50% RWC, the number of amino acids with levels significantly (P < 0.05) greater in S. lepidophylla rose to 12, including increases in tryptophan, N-acetyltryptophan and 3-(4-hydroyphenyl)lactate (Figure 6), as well as 3-ureidoproprionate, N-6-acetyllysine and N-6-trimethylysine (Table S1). γ-Aminobutyrate (GABA), a 4C non-protein amino acid derived from the decarboxylation of glutamate via the GABA shunt (Klähn and Hagemann, 2011), increased significantly (P < 0.05) in abundance in both species at 50% RWC, with abundances more than twofold greater in S. moellendorffii for both hydration states (Figure 6), consistent with its role in stress mitigation (Shulaev et al., 2008). Alanine, tyrosine and 4-hydroxybutyrate (GHB), also part of the GABA shunt, exhibited significantly (P < 0.05) greater accumulation in S. lepidophylla at both hydration states (Figure 6). Several aromatic amino acids, such as tryptophan and its derivative acetyltryptophan, or phenylalanine and its antimicrobial derivative 3-(4-hydroxyphenyl) lactate (Jia et al., 2010), accumulated significantly, particularly in S. lepidophylla at 50% RWC. These aromatic amino acids are known to be biosynthetic precursors for several primary and secondary metabolites in plants (Facchini et al., 2000), and are also increased in abundance in Sporobolus stapfianus following dehydration (Oliver et al., 2011). Alanine and leucine were initially more abundant in S. lepidophylla than in S. moellendorffii, and increased in abundance following dehydration (Figure 6), a pattern observed in the Sporobolus sister-group comparison (Oliver et al., 2011).

Figure 6.

 Differences in amino acid abundances between Selaginella lepidophylla (SL) and Selaginella moellendorffii (SM).
Box plots are as described in Figure 4. The P and Q values for the comparisons are presented in Table S1.

Betaine (glycine betaine), a well-known osmoprotectant in many organisms, including plants, and a favorite target for engineering abiotic stress tolerance (Chen and Murata, 2011), displayed significantly (P < 0.05) elevated abundances in S. lepidophylla for both hydration states, and in S. moellendorffii at 50% RWC (Figure 6). Betaine is a well-known osmoprotectant in Escherichia coli (Miller and Ingram, 2007) as well as in plants (Chen and Murata, 2011). Carnitine and acetyl carnitine also showed an increased abundance in S. lepidophylla relative to S. moellendorffii (Table S1). Carnitine is known to be an osmoprotectant in E. coli (Verheul et al., 1998; Cánovas et al., 2007).

Most nitrogen-rich amino acids (e.g. asparagine, aspartate, arginine and glutamine) were more abundant in S. moellendorffii than in S. lepidophylla for both hydration states (Figure 7). This is in contrast to Sporobolus stapfianus (DT), in which nitrogen-rich amino acids (e.g. asparagine and glutamine) were more abundant than in Sporobolus pyramidalis (DS). In Sporobolus stapfianus, an increased emphasis on nitrogen metabolism was interpreted as being adaptive for life under the nitrogen-limiting conditions typically encountered by this species, or for the production of protective nitrogenous compounds, such as the antioxidant glutathione (Oliver et al., 2011). However, this does not appear to be the case for S. lepidophylla.

Figure 7.

 Differences in nitrogen-rich amino acid abundance between Selaginella lepidophylla (SL) and Selaginella moellendorffii (SM).
Compounds in red or green indicate greater relative abundances in S. lepidophylla or S. moellendorffii, respectively. Box plots are as described in Figure 4. The P and Q values for the comparisons are presented in Table S1.

Notably, citrulline, a structural analogue to arginine known to accumulate in drought-tolerant wild watermelon (Akashi et al., 2001), accumulated in response to water-deficit stress in both species, but the increases were not significant (Figure 7). Citrulline is a non-protein amino acid that accumulates during drought stress in a drought-tolerant watermelon along with arginine and glutamate (Kawasaki et al., 2000). Citrulline is thought to function as a nitrogen reserve or as a hydroxyl radical scavenger (Akashi et al., 2001).

Glutathione metabolism

Within the glutathione metabolism pathway, remarkable differences were apparent between the two species. Whereas S. moellendorffii accumulated greater relative levels of several intermediates, such as glycine, cysteine, 5-oxoproline and glutamate (Figure 8; Table S1), S. lepidophylla accumulated more oxidized glutathione (GSSG) and γ-glutamyl amino acids, especially at 50% dehydration (Figure 8). For γ-glutamyl amino acid dipeptides, eight of the nine compounds detected were more abundant in S. lepidophylla, and showed significantly (P < 0.05) increased abundance in response to dehydration to 50% RWC (Figure 8). Three compounds, γ-glutamylglutamate, γ-glutamylphenylalanine and γ-glutamylvaline, were significantly (P < 0.05) more abundant in S. lepidophylla at both hydration states. For example, γ-glutamyl-phenylalanine, γ-glutamylvaline and γ-glutamylmethionine exhibited a 61.8-, 17.3- and 12.7-fold increase in abundance, respectively, at 50% RWC.

Figure 8.

 Differences in γ-glutamyl amino acid abundance between Selaginella lepidophylla (SL) and Selaginella moellendorffii (SM).
Compounds in red or green indicate greater relative abundances in S. lepidophylla or S. moellendorffii, respectively. Box plots are as described in Figure 4. The P and Q values for the comparisons are presented in Table S1. AA, amino acid; Cys-Gly, cysteinylglycine; GSH, glutathione (reduced); GSSG, glutathione disulfide (oxidized).

Within the γ-glutamyl amino acid cycle, glutathione (GSH), a tripeptide composed of l-glutamate, cysteine and glycine, in which the peptide bond between glutamate and cysteine involves the γ-carboxyl group of glutamate, is broken down in the presence of free amino acids through the action of γ-glutamyltranspeptidase (also known as γ-glutamyltransferase) to produce cysteinylglycine and γ-glutamyl amino acids (Ohkama-Ohtsu et al., 2008). In animal systems, γ-glutamyl cyclo-transferase converts γ-glutamyl amino acids into 5-oxoproline, releasing the amino acid previously captured in the extracellular space into the cytoplasm (Ohkama-Ohtsu et al., 2008). In Arabidopsis, however, this transmembrane recycling of amino acids is minor, and GSH degradation occurs within the cytoplasm (Ohkama-Ohtsu et al., 2008). In S. lepidophylla, however, transmembrane amino acid recycling might occur in a manner that is similar to that found in animal systems, and also provide a means for nitrogen storage, as has been suggested for Sporobolus stapfianus (Oliver et al., 2011). Moreover, the significant increase in γ-glutamyl amino acids during the dehydration of S. lepidophylla, but not during the dehydration of S. moellendorffii, indicates their potential involvement in the acquisition of DT. Furthermore, γ-glutamyl amino acids accumulate in Tortula ruralis and Sporobolus stapfianus in response to desiccation (Oliver et al., 2011), suggesting their role in DT is evolutionarily well conserved. All of the γ-glutamyl amino acids that accumulate in S. lepidophylla, with the exception of γ-glutamylalanine and γ-glutamylvaline (Figure 8) also accumulate in Sporobolus stapfianus, and γ-glutamylvaline accumulates in T. ruralis (Oliver et al., 2011).

Another metabolite that showed greater, but not significantly (P < 0.05) greater, abundance in S. lepidophylla for both hydration states was the oxidized form of glutathione (GSSG), which plays a key role in γ-glutamyl amino acid biosynthesis (Figure 8). GSH, which exists in two forms – reduced and oxidized – provides a reduction equivalent to the oxidation of unstable reactive oxygen intermediates such as hydrogen peroxide or dehydroascorbate by instantly reacting with a similar molecule to form glutathione disulfide (GSSG) (Mittler, 2002). Therefore, greater GSSG abundance in S. lepidophylla compared with S. moellendorffii suggests more activity against oxidative stress. Curiously, the desiccation-induced accumulation of tocopherols observed in Sporobolus stapfianus (Oliver et al., 2011) was not observed in S. lepidophylla, suggesting that strategies for scavenging ROS differ from species to species.

Secondary metabolites

Several amino acid-derived flavonoids (e.g. apigenin, luteolin and naringenin) and a benzenoid (e.g. vannilate) showed significantly (P < 0.05) greater abundance in S. lepidophylla compared with S. moellendorffii for both hydration states (Figure 9). These results indicated that S. lepidophylla is primed to cope with stressful conditions. In many plants, the phenylpropanoid pathway is known to respond to stress and to mitigate ultraviolet light damage and oxidative stress (Iriti and Faoro, 2009). Phenolic and polyphenolic compounds are also thought to preserve resurrection plant lipid bilayer structure, and to protect membrane lipids from ultraviolet light-induced free radical damage (Hagemann et al., 1997).

Figure 9.

 Differences in secondary metabolite abundances between Selaginella lepidophylla (SL) and Selaginella moellendorffii (SM).
Compounds in red indicate greater relative abundances in S. lepidophylla. Dotted arrows indicate a gap in the pathway. Box plots are as described in Figure 4. The P and Q values for the comparisons are presented in Table S1.

Lipids, phospholipids and fatty acids

Membrane integrity is thought to play an important role in desiccation tolerance (Hoekstra et al., 2001). Therefore, several obvious differences in the relative abundances of membrane lipids were noted between the two species. For example, one mono- and four polyunsaturated fatty acids were significantly (P < 0.05) more abundant in S. lepidophylla in both hydration states (Figure 10; Table S1). These changes probably reflect an ability to increase membrane fluidity by altering unsaturated fatty acid levels (Upchurch, 2008). For example, the overexpression of two fatty acid desaturases in tobacco increased tolerance to drought and osmotic stresses, coincident with increased linolenic acid levels (Zhang et al., 2005).

Figure 10.

 Differences in unsaturated fatty acid abundances between Selaginella lepidophylla (SL) and Selaginella moellendorffii (SM).
Box plots are as described in Figure 4. The P and Q values for the comparisons are presented in Table S1.

Several lipoxygenase (LOX) activity markers were greater in S. lepidophylla, especially at 50% RWC (Figure 11). For example, 2-hydroxypalmitate exhibited a 30.7- and 64.3-fold increase at 100 and 50% RWC, respectively. The higher content of these lipid peroxidation products might have resulted from ROS attack or the action of LOXs (Marin et al., 1998).

Figure 11.

 Differences in markers of lipoxygenase activity between Selaginella lepidophylla (SL) and Selaginella moellendorffii (SM).
Box plots are as described in Figure 4. The P and Q values for the comparisons are presented in Table S1. 13-HODE, 13S-hydroxy-9Z, 11E-octadecadienoic acid.

Lastly, with the exception of 1-linoleoylglycerophosphoinositol and 1-oleoylglycerophosphoinositol, which were more abundant in S. lepidophylla, all other phospholipids identified were significantly (P < 0.05) more abundant in S. moellendorffii at both hydration states (Figure 12; Table S1). The triose sugar backbone for glycerophospholipids, glycerol 3-phosphate, also showed a similar pattern of accumulation. The functional significance of these differences in phospholipid compositions remains unclear, but these lipids present excellent candidates for detailed hypothesis testing through future studies.

Figure 12.

 Differences in phospholipids between Selaginella lepidophylla (SL) and Selaginella moellendorffii (SM).
Box plots are as described in Figure 4. The P and Q values for the comparisons are presented in Table S1.

The unknown metabolome

Selaginella moellendorffii displayed more than twice the number of unnamed metabolites of greater abundance than S. lepidophylla in both hydration states. Of the 117 unnamed compounds identified in S. moellendorffii, 25 (21.3%) and 31 (26.5%) were significantly (P < 0.05) more abundant at 100 and 50% RWC, respectively. However, none of these metabolites were more than 0.83-fold more abundant in S. moellendorffii than in S. lepidophylla (Table S1). In contrast, of the 102 unnamed metabolites identified in S. moellendorffii, 65 (63.7%) and 60 (58.8%) compounds were significantly (P < 0.05) more abundant at 100 and 50% RWC, respectively. Of these, 20 (19.6%) unnamed compounds exhibited significantly (P < 0.05) greater abundance in S. lepidophylla at both RWCs, whereas S. moellendorffii contained 54 (46.1%) such compounds (Table S1). However, 14 unnamed metabolites showed between 5- and 120-fold greater abundances in S. lepidophylla than in S. moellendorffii for both hydration states, which suggests that these metabolites could play key roles in DT (Figure 13; Table S1). In fact, five out of the 10 most differentially abundant compounds between the two species were unnamed. At full hydration, X-15003, X-11914 and X-15004 showed ∼121-fold, ∼83-fold and ∼64-fold greater abundances, respectively, in S. lepidophylla than in S. moellendorffii. At 50% RWC, these compounds remained abundant or increased in abundance, with X-11914, X-15003 and X-15006 maintaining ∼113-fold, ∼108- and ∼55-fold greater abundances, respectively, in S. lepidophylla relative to S. moellendorffii. Both species exhibited similar numbers of unnamed compounds (19 in S. moellendorffii; 22 in S. lepidophylla) that increased in abundance following the shift from 100 to 50% RWC (Table S1). A top priority for future research will be to identify this set of 14 unnamed compounds and elucidate their potential roles in the acquisition of DT. These metabolites might also provide exciting targets for metabolic engineering strategies aimed at improving drought tolerance in crop species.

Figure 13.

 Differences in the most abundant unnamed metabolites in Selaginella lepidophylla (SL) and Selaginella moellendorffii (SM).
Box plots are as described in Figure 4. The P and Q values for the comparisons are presented in Table S1.


This sister-group comparison between S. lepidophylla and S. moellendorffii at 100 and 50% RWC provides a global, high-resolution account of the metabolic basis of DT. S. moellendorffii mounts a weak metabolic response to dehydration stress, exhibiting only minor increases in myo-inositol, inositol-1-phosphate and mannitol. In contrast, S. lepidophylla is predisposed to tolerate desiccation in a constitutive manner, using a wide range of metabolites together with some inducible components. Although S. moellendorffii accumulates greater levels of trehalose than S. lepidophylla, this disaccharide is unlikely to be required for the DT of lycophytes. S. lepidophylla constitutively accumulated significantly higher relative levels of sucrose, mono- and polysaccharides, and sugar alcohols than did S. moellendorffii. Other well-known osmoprotectants, such as betaine, and lesser-known osmoprotectants, such as carnitine, showed a greater relative accumulation in the desiccation-tolerant species. Aromatic amino acids and flavonoids were also more abundant in S. lepidophylla, likely playing essential roles in protecting the plants from photo-oxidative and oxidative damage during dehydration. A group of γ-glutamyl amino acids were markedly more abundant in S. lepidophylla, indicating potentially critical roles in the acquisition of DT, perhaps to detoxify ROS via glutathione metabolism and/or remobilize nitrogen following rehydration. Polyunsaturated fatty acids and markers of lipoxygenase activity were greater in S. lepidophylla, especially during dehydration, indicating their potential importance in membrane protection. Lastly, S. lepidophylla contained 14 unnamed compounds with fivefold or greater abundance than in S. moellendorffii, suggesting that these compounds might play critical roles in DT.

Experimental Procedures

Plant material and water deficit stress treatments

Selaginella lepidophylla (Hooker & Greville) Spring (flower of stone) plants were purchased from Hirt’s Gardens ( and maintained in the dried state until use. S. lepidophylla plants were put through a rehydration–dehydration cycle before use in order to identify and remove non-viable tissue. S. moellendorffii (perennial gemmiferous spikemoss) plants were purchased from Plants Delight Nursery Inc. ( and maintained in one-gallon pots containing Metromix 200 (Sun Gro Horticulture Inc., under glasshouse conditions. Natural daylight in the glasshouse was supplemented with high-pressure sodium lamps providing a photon flux density of 200 μmoles m−2 sec−1 on a 14-h light (26°C)/10-h (18°C) dark cycle. To establish the relative water content (RWC) at which S. moellendorffii died, approximately 1-year-old plants were submerged in distilled water, removed from water, blotted dry with an absorbent pad, weighed and placed on a dry absorbent pad in a growth chamber under constant (175 μmoles m−2 sec−1) cool-white fluorescent and incandescent light at 26°C and 35% relative humidity (RH). Six plants were removed at each RWC and replanted to 1-gallon pots containing moist Metromix 200. Plants were moved to glasshouse conditions (described above) and irrigated by misting immediately after all drying treatments, and every 12 h thereafter. Survival rates were scored after 5 days of recovery (Figure S1).

To establish the rate of RWC loss during the dehydration process, nine approximately 1-year-old plants of each species were submerged in distilled water in a growth chamber under constant (175 μmoles m−2 sec−1) cool-white fluorescent and incandescent light at 26°C and 37% RH. Fully hydrated plants were then removed from water and weighed (time zero). After the initial weighing, the plants were blotted dry and placed in dry trays and weighed at regular intervals over a 24-h period, followed by incubation at 65°C for 2 days to obtain dry weights (Figure 1). To establish the rate of RWC gain during rehydration, nine S. lepidophylla plants were air-dried, whereas nine S. moellendorffii plants were dried to 40% RWC and submerged in distilled water before being incubated under the growth chamber conditions described above. Plants were removed from water, blotted dry with an absorbent pad and reweighed at regular intervals (Figure S2). The RWC was calculated according to the formula RWC (%) = (Fwt – Dwt)/(FTwt – Dwt) × 100, where Fwt is the fresh weight at any time point during the dehydration/rehydration cycle, Dwt is the weight after incubation at 65°C for 2 days and FTwt is the weight after 24 h of rehydration. Two RWCs, 100 and 50%, were selected to study the comparative metabolite composition between the two species.

Metabolomic profiling platform

Global unbiased metabolic profiling in S. lepidophylla and S. moellendorffii was performed using three independent approaches: ultra-high performance liquid chromatography/tandem mass spectrometry (UHLC/MS/MS2), optimized for basic species; UHLC/MS/MS2, optimized for acidic species; and gas chromatography/mass spectrometry (GC/MS) (Evans et al., 2009).

Six biological replicates from each species were collected at each time point, lyophilized and kept at −80°C under hermetic conditions prior to extraction. Then, 20 mg of each lyophilized leaf sample was extracted using an automated MicroLab STAR® system (Hamilton Company, in 400 μl of methanol, containing recovery standards. UHLC/MS/MS2 analysis was based on a Waters Acquity UPLC (Waters Corporation, and a Thermo-Finnigan LTQ mass spectrometer (Thermo Fisher Scientific Inc.,, equipped with an electrospray ionization (ESI) source and linear ion-trap (LIT) mass analyzer. Two independent UHPLC/MS injections were performed on each sample using separate dedicated columns: one optimized for positive ions and one for negative ions. Briefly, the chromatographic column used was a 2.1- × 100-mm Waters BEH C18 1.7-μm particle column heated to 40°C. The acidic samples were reconstituted in formic acid and gradient eluted at 350 μl min−1 using: (i) 0.1% formic acid in water; and (ii) 0.1% formic acid in methanol (0–70% in 4 min; 70–98% in 0.5 min; 98% for 0.9 min). The basic samples reconstituted in ammonium bicarbonate were eluted in: (i) 6.5 mm ammonium bicarbonate in water, pH 8.0; and (ii) 6.5 mm ammonium bicarbonate in 95% methanol/5% water (v/v) using the same gradient profile as above, also at 350 μl min−1, as described previously (Evans et al., 2009). For ESI the spray voltages were 4.5 and 3.75 kV for positive and negative ion injection, respectively (Evans et al., 2009).

Retention time variability was assessed and controlled for by using a series of retention time (RT) markers (13 and 11 retention markers on the liquid-chromatography (LC) positive and negative modes, respectively; Evans et al., 2009). These retention time markers were spiked into every sample analyzed, and provided a fixed retention index (RI) that remained constant. Therefore, if retention times drifted as a result of systematic errors, the RI for these markers would not vary. The retention times of the experimentally detected compounds were determined assuming a linear fit between their individual flanking markers. Thus, the RI for each compound was based on its elution relationship to its two surrounding retention markers. The instrument was tuned and calibrated for mass resolution and mass accuracy on a daily basis. Chromatographic separation, followed by full-scan mass spectra, was carried out to record retention time, molecular weight (m/z) and MS/MS2 of all detectable ions present in the samples (Table S2), in compliance with recommendations for reporting metabolite data (Fernie et al., 2011).

The samples destined for GC/MS analysis were re-dried under vacuum desiccation for a minimum of 24 h prior to being derivatized under dried nitrogen using bistrimethyl-silyl-triflouroacetamide (BSTFA). Derivatized samples were analyzed on a Thermo-Finnigan Trace DSQ fast-scanning single-quadrupole mass spectrometer using electron impact ionization, operated at unit mass resolving power. The GC column was 5% phenyl and the temperature ramp was from 40 to 300°C in a 16-min period. For GC/MS analysis, a set of eight retention time standards was used that consisted of a series of alkyl-benzenes, ranging from pentyl-bezene to octadecyl-benzene (i.e. 5, 6, 8, 10, 12, 14, 16 and 18).

A reference library of chemical standard entries that included retention time, molecular weight (m/z), preferred adducts, and in-source fragments, as well as their associated MS/MS2 spectra, was used to rapidly identify metabolites with high confidence using an automated comparison of the ion features in the experimental samples. Artifactual peaks were removed by comparing the experimental samples with process blanks (water only) and solvent blanks. General biochemical pathways, as illustrated in the Kyoto Encyclopedia of Genes and Genomes (KEGG) ( and the Plant Metabolic Network (PMN) (, were used to assign named metabolites to specific metabolic networks.

Data imputation and statistical analysis

In order to minimize process variance, samples in this study were analyzed over the course of 2 days. Overall process variance, which includes MS peak quantitation variance, was established by extracting and running a set of five or six technical replicates, interspersed evenly among the experimental samples throughout a run day. The technical replicates were made from a pool of experimental samples, so they represent the average matrix being studied. Minor variations resulting from between-day instrument tuning differences were corrected. If missing values for a given metabolite were detected, then such values were assigned the observed minimum detection value. This procedure was based on the assumption that the missing values were below the limits of detection (Evans et al., 2009). Raw area counts for each biochemical were re-scaled by dividing each sample value by the median value for the specific biochemical in order to visualize the data more conveniently.

jmp (SAS,, a commercial software package, and r (, a freely available open-source software package, were used for the statistical analysis of the data. A log transformation was applied to the observed relative abundances for each biochemical because the variance generally increased as a function of the average response for each biochemical.

A total of 199 metabolites with no missing values were imported into the chemometrics software solo (Eigenvector Research Inc., for partial least-squares discriminant analysis (PLS-DA) to determine the classification of the treatments (Figure 3). R2 and Q2 are used as measures for the robustness of a PLS-DA model, where R2 is the fraction of variance explained by a model. The cross-validation of R2 estimates Q2, which explains the fraction of the total variation predicted by the model.

To visualize the entire data set, a heat map was generated to show the fold change for each compound identified from GC-MS and LC-MS analyses of the tissue samples (Table S1). Fold change was calculated for each compound identified as the means ratio of each treatment in S. lepidophylla compared with the same treatment in S. moellendorffii. Welch’s two-sample Student’s t-tests were then used to determine whether or not each metabolite significantly increased or decreased in abundance. The false discovery rate (FDR) was then calculated to correct for multiple Welch’s two-sample Student’s t-test comparisons for the hundreds of compounds detected. Similar to gene array studies, metabolomic profiling generates large numbers of metabolites; hence, FDR was used for multiple comparison adjustment. The FDR for a given set of metabolites is estimated by the Q-value (Storey, 2002). Box plots were generated for compounds that showed a significant increase or decrease using both Welch’s two-sample Student’s t-test and FDR (i.e. P < 0.05 and Q < 0.10) significance values.


This work was supported, in part, by a grant from the USDA National Institute of Food and Agriculture to MJO and JCC (grant no. 2007-55100-18374), and by Hatch funding from the Nevada Agricultural Experiment Station (NAES-0341). The authors would also like to thank Mary Ann Cushman for her helpful and clarifying comments on the article. Mention of a trademark or proprietary product does not constitute a guarantee or warranty of the product by the United States Department of Agriculture, and does not imply its approval to the exclusion of other products that may also be suitable. The authors have no conflicts of interest to declare.