Concept and Experimental Design

We used a systems approach to elucidate the genes and proteins that are induced or activated in response to desiccation stress. The outcome of this analysis was used to define candidate anhydrobiotic pathways, which were validated by a standardized desiccation tolerance assay on mutant worms or worms treated with RNAi.

First, the dauer larva should have a mechanism that senses a decrease in ambient humidity, namely hygrosensation. The signal of a desiccative environment, regardless of how it is detected, should be transformed into a response that is manifested at all levels of gene expression: transcriptional, translational, and post-translational. Figure S1A shows possible scenarios of this response in the C. elegans dauer larva. Desiccation tolerance relevant (DTR) mRNAs or proteins could be produced during dauer formation. Upon preconditioning, these DTR mRNAs can be activated and additional desiccation-induced DTR mRNAs can be synthesized de novo. Both types of DTR mRNAs can lead to de novo synthesis of DTR proteins. Hygrosensation can also induce post-translational modifications (e.g., phosphorylation or glycosylation) of these proteins. DTR proteins will then execute the functions required for anhydrobiosis.

To explore these scenarios, we analyzed the differential expression of genes at the mRNA level and the corresponding changes at the proteome level, as well as the post-translational modifications of proteins. For this purpose, we used microarray analysis, gel electrophoresis liquid chromatography tandem mass spectrometry (geLC-MS/MS), and 2-dimensional difference gel electrophoresis (2D-DIGE) (Figure S1B). A temperature-sensitive dauer-constitutive strain of C. elegans, daf-2(e1370), was grown in liquid culture at 25 °C until complete dauer formation [19]. One aliquot of these worms was immediately frozen before any desiccation, and was thus treated as a non-desiccated control. The rest were subjected to mild desiccation at 98% RH for 1 day or 4 days at 25 °C (preconditioning). Dauer larvae were collected and total RNA and protein were extracted for transcriptome and proteome analysis, respectively (Figure S1B, upper part). To test the reversibility of changes in the proteome during preconditioning, an aliquot of the 4-day-desiccated worms was kept in water for 24 h to rehydrate. Subsequently, proteins were also extracted from this sample.

Various bioinformatics tools were used to identify candidate genes and pathways that were differentially expressed during preconditioning, and thus presumably have a function in the desiccation response. Finally, to assign functions to these pathways, we tested the desiccation tolerance of mutant worms deficient in genes of the pathway or used an RNAi knockdown approach (Figure S1B, lower part).

Microarray Analysis of Genes That Are Induced Upon Desiccation

Using microarray technology, we surveyed the differential expression of 20,058 protein-coding genes of C. elegans in the dauer larva before and after preconditioning. During the first day of exposure to decreased RH, 4,764 (23.7%) of these genes were significantly upregulated and 4,791 (23.9%) were significantly downregulated (p < 0.001). However, the majority of these genes had very small differential expression levels (fold changes, FCs), and were therefore regarded as being biologically non-relevant. To filter these genes out without using arbitrary FC cutoffs, we used k-means clustering on FC values. This is one of the simplest unsupervised machine learning algorithms that divides a dataset into a predefined number of subsets by optimizing the distance between means. Using this algorithm, we grouped all significantly upregulated and downregulated genes into four sets: very low, low, medium, and high FC clusters (FCCs) (Figure 1A). The very low FCC revealed by k-means clustering (less than 1.82-fold upregulation and 1.92-fold downregulation) was excluded from further analyses. Finally, 1,833 (9.1%) upregulated and 2,433 (12.1%) downregulated genes remained (Dataset S1). These genes were considered to be statistically significant and biologically relevant.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Analysis of desiccation-induced DTR mRNAs.

(A) Fold change clustering of upregulated (green area) and downregulated (pink area) genes. The very low FCC was discarded because the differential expression was too low. The number of genes (n) in each group is shown. (B) Hierarchical clustering of the 64 highly upregulated genes (blue region in panel A). Red/green column shows log₂ expression levels before (–) and after (+) preconditioning, cyan column shows log₂ fold changes for the genes indicated on the right. Color codes for the heat map are shown top left. Main branches of the dendrogram are labeled ‘a’ (light gray) and ‘b’ (dark gray, see the text for details). Highlighted genes reside in essential anhydrobiotic pathways shown in the legend with unique color codes, which are consistent in the following figures. (C) The trehalose biosynthesis pathway is upregulated in C. elegans upon desiccation stress. Enzymes that catalyze each reaction are shown with the corresponding differential expression values of their transcripts upon preconditioning. These fold changes are shown in green, red, and blue for upregulation, downregulation, and no change, respectively. Glucose-6-phosphate synthesis from lipids involves several steps; therefore, a dashed arrow is used. Highlighted genes are found in the high FCC (panel B). Glc, Glucose; Glc-6-P, glucose-6-phosphate; Glc-1-P, glucose-1-phosphate; UDP-Glc, UDP-glucose; Tre-6-P, trehalose-6-phosphate; Tre, trehalose.

https://doi.org/10.1371/journal.pone.0082473.g001

Most interesting for further analysis were two FCCs: medium (444 genes with 3.5–13 FC) and high (64 genes with at least 13 FC) (Figure 1A). The top 10 genes were upregulated by more than 100-fold upon preconditioning. We considered the genes in the high FCC as major candidates to have specific functions in desiccation tolerance. Therefore, they were expected to be expressed at relatively low levels before preconditioning. We tested this hypothesis by clustering these genes according to their expression levels in non-preconditioned and preconditioned samples (Figure 1B). Here, we took advantage of the one-color microarray hybridization technology in which the signal intensity of a probe correlates with the abundance of its target mRNA molecule, i.e., its expression level [27]. Hierarchical clustering of the control and preconditioned expression levels of high FCC genes resulted in two patterns: about half of them were expressed at low levels before preconditioning and their transcript levels increased enormously upon desiccation stress (Figure 1B, branch a). The remaining genes were expressed at a medium level before preconditioning and were induced further upon desiccation, often eventually being expressed at the highest levels (Figure 1B, branch b). Thus, genes of the first pattern appear to be specific for the desiccation response, whereas genes of the second pattern might also be involved in the regular functioning of the dauer and be required prior to desiccation. Alternatively, they might be necessary to provide a basal level of desiccation tolerance to the dauer, which is increased upon desiccation stress by the activation of DTR genes and proteins.

One possible way to analyze the desiccation stress response would be to investigate the roles of individual genes identified by microarray analysis. However, many C. elegans genes are functionally highly redundant [28], which would make such analysis extremely difficult, if not impossible. Therefore, we first assigned the identified candidates to biochemical and genetic pathways and then investigated the involvement of the whole pathway in desiccation tolerance. As a proof of principle, we tested whether the strong increase of trehalose during preconditioning, which we observed in our previous study [19], was reflected in the current microarray data. The biosynthetic pathway of trehalose and the genes involved in this pathway are shown in Figure 1C. Worms synthesize trehalose in a two-step reaction from glucose-6-phosphate and UDP-glucose. The first step is catalyzed by trehalose-6-phosphate synthase, which is encoded by two paralogous genes in C. elegans: tps-1 and tps-2 [29]. Trehalose-6-phosphate is then dephosphorylated by trehalose-6-phosphate phosphatase, yielding trehalose [30]. Several enzymes catalyze the production or conversion of the substrates required for TPS-1 and TPS-2 (note the genetic redundancy of the enzymes in the pathway). As seen in Figure 1C, all the genes encoding the enzymes involved in trehalose production were upregulated upon preconditioning with the exception of an isoform of hexokinase, F14B4.2, which remained unchanged. Most importantly, the genes encoding two enzymes (phosphoglucomutase R05F9.6 and UDP-glucose pyrophosphorylase D1005.2) were found in the high FC cluster (highlighted in brown in panels B and C). In addition, three of the five trehalose hydrolyzing enzyme (trehalase) genes were downregulated. These data clearly show that microarray data can be used to predict the pathways involved in the desiccation response in worms.

Next, we performed functional annotation clustering of differentially expressed genes based on their Gene Ontology and conserved protein domain annotations (Dataset S2, summarized in Table S1) [31,32]. These clustering results suggested that activation of many enzymatic processes participates in lipid and sugar metabolism. Especially, upregulation of lipid-binding protein, lipase, fatty acid (FA) desaturase (FAT), and lipid glycosylation enzyme genes suggests that lipidome remodeling can be intensive during preconditioning. At the same time, transcriptional activation of genes involved in mitochondrial substrate/solute carrier activity could be an indication of elevated energy production during preconditioning, at least in the early stages. The data also indicated strong activation of the ROS defense pathway. Additionally, many enzymes with cytochrome P450 and glutathione S-transferase (GST) activity were enriched among the upregulated genes. These enzymes not only accept endogenous substrates but also contribute to degradation of xenobiotics. Thus, detoxification may be another general strategy of desiccation tolerance.

Overall, it seems that during entry into anhydrobiosis, worms reduce or cease building new morphological structures (e.g., cuticle) while preserving the existing ones. This is evident from the downregulation of cuticle and extracellular matrix proteins, lipid glycosylation enzymes, and proteases, as well as lysozymes that break down peptidoglycans. Amino acid metabolism and transport are also reduced, which might be an indication of the onset of metabolic depression in conjunction with decreased lipase activity (Table S1).

The C. elegans Dauer Larva Responds to Desiccation at the Translational and Post-translational Levels

To survey the desiccation-induced changes at the proteome level, we used two approaches: a label-free protein quantification method geLC-MS/MS [33] and a minimal-labeling method 2D-DIGE. geLC-MS/MS detected 1,058 proteins expressed in the C. elegans dauer. However, the relative abundance of the majority of these proteins did not significantly change during desiccation. After 1 day of desiccation stress, the abundance of only 17 (1.6%) proteins was significantly increased. On the fourth day, the levels of 12 of these 17 proteins were further increased or whereas 5 remained unchanged. Additionally, the abundance of 22 other proteins increased and the abundance of nine proteins decreased, resulting in a total of 39 (3.7%) upregulated and 9 (0.9%) downregulated proteins (Dataset S3). These results suggest that desiccation-induced protein expression is very specific. It also indicates that desiccation mainly induces protein biosynthesis rather than protein degradation, as very few downregulated proteins were detected.

Proteins whose abundance was most strongly altered during preconditioning were mainly ROS defense and detoxification enzymes, FA- and retinol-binding proteins, HSPs, and intrinsically disordered proteins (IDPs), as well as transthyretin-like proteins. These classes correlated quite well with the data obtained from the microarray analysis of DTR mRNAs that are upregulated during preconditioning. The first 22 proteins with at least 2 FC, or their paralogs, were also found in the medium or high FCCs of the microarray survey. The only exception was the uncharacterized protein F57H12.6, which was found exclusively by the geLC-MS/MS approach. Its expression remains unchanged according to the microarray data.

Another powerful, unbiased tool to identify differential expression at the protein level is 2D-DIGE. In comparison to geLC-MS/MS, this method has notable benefits in revealing changes in the patterns of post-translational modifications of proteins. 2D-DIGE of protein extracts from non-desiccated and 4-day-desiccated dauer larvae is shown in Figure 2 (red and green spots, respectively). The abundance of the majority of proteins did not change during desiccation (yellow spots), which corroborates the microarray and geLC-MS/MS findings. Remarkable de novo synthesis of several isoforms of LEA-1, an intrinsically disordered protein, was noted in the desiccated samples (Figures 2, S2B, S2C). The implications of this are discussed below.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Comparison of proteomes upon desiccation.

Overlay of false-colored 2D-DIGE images comparing dauer proteomes before (red) and after (green) preconditioning at 98% RH for 4 days. Yellow spots indicate proteins that do not change distinctively. The major proteins identified in these spots are annotated. The numbered regions are shown in higher magnification in Figures S2C–E. For MLC-1/2 (region 8), the red and green spots were identified as phosphorylated and dephosphorylated proteins, respectively. Region 9 possibly shows another post-translational modification.

https://doi.org/10.1371/journal.pone.0082473.g002

Well known as a stress-resistant stage, the dauer larva is remarkably different from its reproductive stage counterpart L3 larva. These differences might also contribute to the desiccation tolerance. We addressed this hypothesis by comparing the proteomes of L3 and dauer larvae using 2D-DIGE (Figure S2A). Clearly, the two stages differ dramatically at the protein level. While some proteins such as SOD-3 are also expressed in L3 larvae, although at lower levels, others such as DUR-1, CTL-1, and CTL-3 are entirely dauer-specific (Figures S2A, S2D). This indicates that the transition to dauer might confer partial desiccation tolerance ability, and that full desiccation tolerance is acquired upon desiccation stress.

In addition to protein expression, desiccation also induces post-translational modifications to existing proteins. Phosphorylation of actin and dephosphorylation of myosin light chain proteins were detected upon desiccation (Figures 2, S2E).

Interestingly, upon rehydration, many DTR proteins were downregulated to their basal levels although the levels of some proteins remained elevated (Figures S2B, S2C). Desiccation-induced post-translational modifications were also reversed in many cases upon rehydration (Figures S2B, S2E).

Data obtained by geLC-MS/MS and 2D-DIGE complement those of the microarray survey and thus were incorporated into efforts to identify the pathways responsible for desiccation tolerance.

Major Biochemical Pathways Underlying Desiccation Tolerance

Unbiased bioinformatics examination of the transcriptome and proteome data led to the identification of candidate pathways that were induced by desiccation. Nevertheless, subsequent biological validation was essential to associate these pathways with anhydrobiosis. Therefore, we tested selected genes in some of these pathways for their involvement in desiccation tolerance (Table 1). Dauer larvae carrying mutations in the listed genes were first preconditioned at 98% RH for 4 days. Then they were either kept at 98% RH for another day or subjected to a harsher desiccation at 60% RH, which cannot be tolerated in the absence of essential DTR genes such as tps-1 and tps-2 [19]. Finally, the percentage of survivors was calculated by counting the worms. When mutant strains were unavailable, we silenced the candidate gene by RNAi on daf-2 background and performed the same desiccation assay. In these experiments, daf-2 eggs were grown on RNAi feeding plates at 25 °C for one generation until they form dauer larvae. Every mutant or RNAi knockdown was tested at least twice. In every desiccation experiment, wild type (N2) or RNAi non-treated daf-2 worms were used as the control.

Survival data were analyzed using beta regression [34]. The estimated survival rates and their approximate standard errors are presented in Table S2. We categorized the desiccation sensitivity phenotype of mutants according to their survival rates into 4 groups. Any strain whose survival rate was not significantly different from the control (p ≥ 0.05) was considered as desiccation tolerant. The others were categorized as desiccation sensitive (> 50% survival), very sensitive (25 - 50% survival) and extremely sensitive (< 25% survival).

A. Small HSPs Are Essential for Desiccation Tolerance

According to microarray and proteomics surveys, several HSPs were highly upregulated upon desiccation stress (Datasets S1, S3). The gene with the overall highest differential expression (704-fold), F08H9.4, encodes an HSP-16-like protein.

Various types of abiotic stress may cause protein aggregation, which is eliminated by the cell via molecular chaperones known as HSPs, and by proteases when necessary. HSP90, HSP70, and chaperonins act as molecular chaperones via ATP hydrolysis [35], whereas small HSPs and α-crystallins bind proteins in non-native conformations in an ATP-independent manner [36]. Certain HSPs were first implicated in plant desiccation tolerance [37], then also in the anhydrobiotic African chironomid [18], a tardigrade [38], and a nematode [39].

F08H9.3, F08H9.4 and hsp-70 mutants were desiccation sensitive (Figure 3A). A particularly strong phenotype was observed in the HSP-16-like protein mutant F08H9.3, which was already affected during preconditioning. Remarkably, despite the redundancy of genes in this class, disruption of a single HSP gene still induced a strong phenotype. We conclude that HSPs, especially small HSPs, are essential for desiccation tolerance in C. elegans.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Major pathways involved in desiccation tolerance.

(A) Desiccation sensitivities of various mutants from (B-D) candidate pathways are presented in 4 categories. Genes from high FCC are highlighted according to Figure 1. The subcellular localizations of ROS defense enzymes are shown as c, cytosol; e, extracellular; m, mitochondrion; and p, peroxisome.

https://doi.org/10.1371/journal.pone.0082473.g003

B. Desiccation Response via LEA-like Protein Expression

LEA-1, a C. elegans homolog of plant LEA proteins, was the major protein whose abundance increased upon preconditioning in 2D-DIGE gels (Figure 2, S2C). Microarray and geLC-MS/MS approaches also detected LEA-1 as a gene/protein that was highly expressed during desiccation (Datasets S1, S3). LEA proteins are associated with desiccation tolerance in plant seeds, as well as the vegetative tissues of anhydrobiotic plants. Recently, LEA homologs were discovered in many animal species [17]. Classified as IDPs, LEA proteins obtain their functionally active conformations upon desiccation, unlike other proteins [40]. They can reduce protein aggregation by lessening the desiccation-induced interactions between proteins and polypeptides, therefore acting as molecular shields [41,42].

Because the lea-1 mutant is not available, we performed the desiccation assay using RNAi. lea-1-knockdown worms were very sensitive to desiccation at 60% RH but not at 98% RH, unlike it was previously reported [43] (Figure 3A). Given that RNAi seldom fully eliminates the target mRNA, it can be concluded that LEA-1 protein is essential for desiccation tolerance.

In addition to lea-1, C. elegans possesses another gene, dur-1, which encodes a homolog of rotifer (Adineta ricciae and Adineta vaga) LEA proteins (Figure S3). DUR-1 is also similar to C. elegans LEA isoforms, as well as LEA isoforms from two other nematodes (Caenorhabditis remanei and Caenorhabditis briggsae) and some plants (Medicage truncatula and Arabidopsis lyrata lyrata). dur-1 mutants were extremely sensitive to desiccation at 60% RH (Figure 3A). Furthermore, these mutants were already very sensitive to mild desiccation, which suggests that DUR-1 is required already during preconditioning. Altogether, these results clearly indicate that IDPs are essential for the anhydrobiotic response.

C. Desiccation Tolerance Requires Cytoplasmic ROS Defense

Another prominent pathway revealed by the microarray and proteome data was ROS defense. Biochemical reactions that use electron transfer to molecular oxygen (such as respiration and photosynthesis) naturally cause the formation of ROS. These are highly reactive molecules, which can damage proteins, lipids, and DNA inside cells. Therefore, cells have enzymatic and non-enzymatic mechanisms to prevent the formation of ROS or to repair the damage caused by them. A major ROS is the superoxide radical (O₂^–), which is produced by the transfer of a single electron to molecular oxygen (Figure 3B). Superoxide dismutases (SODs) can convert O₂^– into hydrogen peroxide (H₂O₂) and oxygen (O₂). However, H₂O₂ is itself an ROS and should be inactivated. Catalases (CTLs) can directly convert H₂O₂ into H₂O and O₂. Alternatively, glutathione (GSH) peroxidases (GPXs) can convert H₂O₂ into H₂O via reducing GSH [16] (Figure 3B). In addition to SODs, CTLs, and GPXs, worms have peroxiredoxins, glutaredoxins, thioredoxins, and thioredoxin reductases that participate in ROS detoxification. Our microarray survey showed that all SODs, five GPXs, and CTL-2, as well as some peroxiredoxins, glutaredoxins, thioredoxins, and thioredoxin reductases were upregulated upon desiccation stress (Figure 3B, Dataset S1). Two of these genes (sod-5 and gpx-6) were in the high FC class. Furthermore, 2D-DIGE analysis showed that CTL-1 and CTL-3 proteins were strongly induced during the dauer transition and were upregulated by preconditioning (Figures S2A, S2D).

The desiccation assay showed that worms with mutated components of the ROS defense pathway, namely a cytosolic SOD (SOD-1) [44], two peroxisomal GPXs (GPX-2 and GPX-7) [45], a cytosolic GPX (GPX-6), and a cytosolic CTL (CTL-1), were sensitive to desiccation at 60% RH (Figure 3A). However, all of the mutants were desiccation tolerant at mild humidity (98% RH). Similar to the small HSP pathway, ROS defense genes are redundant. Despite this, gpx-7 mutant was extremely desiccation sensitive (Figure 3A). These results indicate that cytosolic and peroxisomal ROS defense have an essential role in desiccation tolerance.

D. Desiccation Induces Detoxification Mechanisms

Among the upregulated genes of the high FCC in the microarray survey were two genes encoding glyoxalases: djr-1.2 and glod-4 (Figure 1B). The main function of glyoxalases is to detoxify α-oxoaldehydes such as glyoxal (GO) and methylglyoxal (MGO), which are byproducts of glycolysis [46] (Figure 3C). These molecules react with the amino groups of proteins and form advanced glycation end-products, and thereby cause protein damage. In mammals, GO and MGO are detoxified by the glyoxalase I/II system in a GSH-dependent manner [47]. In C. elegans, GLOD-4 protein is a glyoxalase I and decreases mitochondrial ROS production [48]. A novel type of glyoxalase homolog, glyoxalase III, was recently discovered in humans, mice, and worms [49]. This enzyme, DJ-1 (also named PARK7 because it has been associated with early-onset Parkinson’s disease [50]) is GSH-independent and can directly detoxify GO and MGO. C. elegans has two genes, djr-1.1 and djr-1.2, that encode DJ-1 isoforms [49]. Deletion of both of these genes on daf-2 background (daf-2;∆∆djr) rendered the worms sensitive to desiccation (Figure 3A). However, single mutants of djr-1.1, djr-1.2 or glod-4 did not exhibit desiccation sensitivity. This is most likely due to the redundant functions of glyoxalase systems. Further investigation of djr-1.1;djr-1.2;glod-4 mutants will help us understand better the role of glyoxalase activity on desiccation tolerance.

Other xenobiotic degradation mechanisms might also be required for anhydrobiosis. Cadmium (Cd) toxicity is a xenobiotic stress for C. elegans. A novel Cd-responsive gene, cdr-1, and its six paralogs (cdr-2–7) have previously been identified [51,52]. Cd treatment induces transcription of cdr-1 by more than 50-fold. However, the other cdr genes are not induced as dramatically as cdr-1 [52]. All these proteins are predicted to be transmembrane [52] and to contain putative GST domains (Table S3). Our microarray survey revealed that the levels of all cdr transcripts, except cdr-1 and cdr-5, significantly increased upon desiccation stress. Mutation of cdr-2 results in extreme desiccation sensitivity at 60% RH, whereas mutation of cdr-3 has a milder effect (Figure 3A). These results suggest that the detoxification function of cdr genes may have a role in desiccation tolerance.

E. Polyamine Biosynthesis as an Anhydrobiotic Strategy in Animals

According to the geLC-MS/MS analysis, the level of the T21F4.1 protein was very strongly elevated during preconditioning (Dataset S3). Furthermore, in the microarray survey, this gene was found in the medium FCC. T21F4.1 is an ortholog of human arginase 1, which is mainly involved in the urea cycle and is partially involved in the biosynthesis of polyamines. Polyamines, namely putrescine, spermidine, and spermine, are polycations that are associated with diverse functions in the cell. As in other organisms, the first step of this pathway in C. elegans is the conversion of arginine into the amino acid ornithine by the mitochondrial enzyme arginase [53] (Figure 3D). Subsequently, ODC-1 produces putrescine by decarboxylation of ornithine [54], which is then converted into spermidine by SPDS-1 [55] (Figure 3D).

We tested available strains bearing mutations in genes of this pathway. Although levels of SPDS-1 and ODC-1 mRNA did not change upon desiccation according to the microarray survey, the effect of odc-1 and spds-1 mutations on the desiccation tolerance of dauer larvae was very strong (Figure 3A). In fact, spds-1 mutants were extremely sensitive to desiccation already at 98% RH. These results suggest that polyamines, especially spermidine, are required for desiccation tolerance.

Recently, polyamine back-conversion (i.e., spermine to putrescine) via polyamine oxidases was implicated in drought stress in plants [56]. Our microarray survey showed a 2.8-fold upregulation of the amine oxidase gene amx-2 (Dataset S1), which has homology to polyamine oxidases from various organisms. Thus, along with spermidine biosynthesis, putrescine back-conversion might be an essential anhydrobiotic strategy used by C. elegans. However, the molecular details of this strategy remain to be investigated.

F. Role of Polyunsaturated Fatty Acids in Desiccation Tolerance

Preconditioning induces several lipid-binding proteins and lipid-modifying enzymes (cf. subsection G below). Among these is the class of FATs, which produce monounsaturated FAs (MUFAs) from dietary or de novo synthesized palmitic acid, and subsequently polyunsaturated fatty acids (PUFAs) from MUFAs (Figure 4A) [57]. The C. elegans genome has seven genes (fat-1–7) that encode five types of FATs. Among these, FAT-5, FAT-6, and FAT-7 can redundantly catalyze ∆9 desaturation [58]. This gives rise to MUFAs, among which oleic acid (C18:1n-9) is further desaturated into linoleic acid (C18:2n-6) by FAT-2. This is the first PUFA and is a precursor to all other PUFAs in subsequent desaturation and elongation reactions [59]. The end product of this pathway is eicosapentaenoic acid (EPA, C20:5n-3), the major PUFA in C. elegans [57,60] (Figure 4A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Arachidonic acid is essential for desiccation tolerance.

(A) The major components in the biosynthesis of MUFAs and PUFAs in C. elegans (highlighted according to Figure 1). Type of desaturation is shown in orange for every reaction. Modified from [57,59]. (B) Desiccation sensitivities of FAT mutants in 4 categories. (C) PUFA profiles of preconditioned dauer larvae of wild-type worms and fat mutants. Filled and empty boxes indicate the presence and absence of the PUFA species, respectively. LA, Linoleic acid; ALA, α-linolenic acid; GLA, γ-linolenic acid; SA, stearidonic acid; DGLA, dihomo-γ-linolenic acid; O3AA, ω-3 arachidonic acid; AA, ω-6 arachidonic acid; EPA, eicosapentaenoic acid.

https://doi.org/10.1371/journal.pone.0082473.g004

Preconditioning induced ∆9 (FAT-5 and FAT-6), ∆12 (FAT-2), and ∆5 (FAT-4) FATs, as well as FA elongases (ELO-1, ELO-2, and ELO-9) (Figure 4A, Dataset S1). We first tested the desiccation tolerances of the ∆9 desaturase mutants fat-5, fat-6, and fat-7. All these mutants were extremely sensitive to desiccation at 60% RH (Figure 4B). In fat-6 and fat-7 mutants, desiccation sensitivity appeared already at 98% RH. Remarkably, single mutants do not have any detectable phenotypes under normal conditions and are superficially wild-type [57]. By contrast, all three ∆9 desaturases are essential during desiccation. An explanation for this could be that FATs are differentially expressed in various cells. Thus, the viability of the cell is not affected in the absence of unsaturated FAs under normal conditions; however, desaturation seems to be essential for desiccation tolerance.

∆9 FA desaturation is the first step in the production of PUFAs. Are PUFAs required for the desiccation tolerance, similar to MUFAs? To address this question, we tested fat-3, fat-4, fat-1, and fat-4;fat-1 (designated as fat-1,4) mutants in the desiccation assay. Except for fat-1, all mutants were extremely desiccation sensitive at 60% RH. (Figure 4B). The same mutants were affected by mild desiccation as well. Thus, we conclude that not only MUFAs but also PUFAs are required for desiccation tolerance.

The schematic in Figure 4C shows the FA contents after preconditioning (as detected by us in the dauer larva, similar to non-preconditioned reproductive stage worms [61]). The correlation between the FA contents and desiccation resistance phenotypes suggests that worms require C20 PUFAs (because the fat-3 mutant was extremely sensitive to desiccation) and, in particular, ω-6 arachidonic acid (AA) (because the fat-4 and fat-1,4 mutants were extremely sensitive, whereas the fat-1 mutant was desiccation tolerant). This suggests that AA is an essential PUFA for desiccation tolerance.

In addition to MUFAs and PUFAs, C. elegans can also produce hydroxy- and epoxy-derivatives of AA and EPA [62], which most probably act in signal transduction. One of the genes involved in this process, cyp-33C9, was significantly upregulated in response to desiccation in our microarray survey. When tested in the desiccation assay, the cyp-33C9 mutant displayed desiccation sensitivity, although this phenotype was much less severe than that of the fat mutants (Table S2). This suggests that not only AA (and possibly EPA) but also their hydroxy- and epoxy-derivatives may be required for desiccation tolerance.

G. Other Novel Strategies for Desiccation Tolerance

In addition to those mentioned above, we chose mutants of other annotated genes that were induced by preconditioning, according to our transcriptome and proteome data. The desiccation tolerances of these mutants were tested (Table 1). Among them, cex-1, cex-2, try-5, ugt-1, and C04G2.2 mutants showed sensitivity to desiccation at 60% RH (Figure 5A). Furthermore, cex-1, cex-2 and ugt-1 mutants were desiccation sensitive already at 98% RH.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. Novel DTR proteins and putative elements of the hygrosensation pathway involved in desiccation tolerance.

https://doi.org/10.1371/journal.pone.0082473.g005

cex-1 and cex-2 are C. elegans homologs of the Ca-binding protein calexcitin, which putatively regulates membrane excitability in neurons [63] and muscle cells [64]. try-5 has recently been identified as an extracellular sperm-activating protease [65]. ugt-1 is a UDP-glucosyltransferase that is genetically activated by daf-16 [66]. Its closest human ortholog, UGT1A6, is involved in phenol detoxification [67]. C04G2.2 is a nematode-specific tau tubulin kinase-like protein [68]. However, knockdown of this gene does not significantly affect the life span of the animal [66].

In addition to novel DTR genes with known function, we found many uncharacterized genes in the high FCC. To deduce the functions of the proteins encoded by these genes in desiccation tolerance, we analyzed their sequences using various bioinformatics tools (Table S4).

Nematodes often display many uncharacterized genes that lack clear orthologs in higher organisms; therefore, we focused on the conserved domains of the proteins of interest. In four proteins from the high FCC (F11F1.4, F11F1.5, F11F1.8, and F13E9.14), a domain of unknown function (DUF148) was found, which seems to be nematode-specific (Figure S4A). Hierarchical clustering of expression levels of high FCC proteins showed that three of these paralogs (F11F1.4, F11F1.5, and F11F1.8) have a very similar differential expression pattern (Figure 1B). Homology detection and structure prediction by HMM-HMM comparison (HHPRED) of DUF148 revealed similarity to a proteobacterial domain (Lipase_chap, PF03280) that is present in lipase chaperones [69] (Figure S4B). These chaperones are required for functionally active lipases [70]. DUF148 lacks part of the N-terminal sequence of the Lipase_chap domain (Figure S4B), which suggests that the molecular functions of DUF148-containing proteins might be different from those of their proteobacterial orthologs. Overall, this finding suggests that desiccation tolerance requires a wide array of chaperones, including HSPs, IDPs, and lipase chaperones. Notably, all four DUF148 proteins contain an N-terminal signal peptide, and thus could be secreted. One of them, F13E9.14, has in addition several YGG-repeats, which are also found in the P granule proteins PGL-1 and GLH-1 in C. elegans embryos [71].

As a next step to understand the possible function of the uncharacterized high FCC proteins, we searched for their potential interacting partners by using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) [72]. Figure S5 shows all identified molecular interactors of four uncharacterized high FCC proteins. Among these, R05D7.2 seems to be associated with the nucleolar RNA processing machinery (Figure S5A). F53A9.2 may be involved in carbohydrate metabolism (Figure S5B) and C40H1.3 is associated with a SUMO ligase (Figure S5C). Finally, C54F6.5 and F41C3.1 are connected to the tumor suppressor and E3 ubiquitin ligase VHL-1, as well as a cytochrome P450 (Figure S5D).

As mentioned before, desiccation stress might induce post-translational modifications of proteins. Via mass spectrometric analysis of gel-isolated spots, we found that myosin light chain proteins MLC-1 and MLC-2 were dephosphorylated during preconditioning (Figures 2, S2E). This modification was reversed by rehydration (Figures S2B, S2E). Furthermore, the actin isoform ACT-4 was possibly phosphorylated during preconditioning (Figures 2, S2E) and rehydration (Figure S2E). These data imply that certain post-translational modifications of cytoskeletal proteins may be required in the anhydrobiotic state. Therefore, cytoskeletal reorganization could be another molecular strategy of anhydrobiosis in C. elegans.

The Desiccative Environment Might Be Sensed by Head Neurons

All our data indicate that mild desiccation stress (preconditioning) induces differential expression of many genes at the mRNA and protein levels. In this way, the organism is prepared for an anticipated harsher desiccation condition. Thus, dauer larvae should be able to detect changes in humidity and have mechanisms to convert the desiccation signals into changes at the protein expression level.

Sensing of the osmolarity of the environment has been thoroughly studied in C. elegans [73-76]. We speculated that sensing of humidity could involve similar genes and molecular mechanisms. Indeed, it was gratifying to find that osm-11 and its paralog dos-1 were significantly induced upon desiccation stress (Datasets S1 and S3). osm-11 is one of the osm genes (OSMotic avoidance abnormal) that are required for osmotic avoidance [76]. OSM-11 is a Notch ligand in seam cells and acts on Notch receptors in head neurons [77]. Together, OSM-11 and DOS-1 are required for avoidance of 1-octanol [77]. osm-11 mutants were very sensitive to desiccation at 60% RH (Figure 5B). This suggests that hygrosensation might be associated with certain head neurons. The role of one such neuron, ASH, in osmosensation is well established [74]. ASH neurons express TRP channels formed by OSM-9 and OCR-2, which are associated with various sensory functions [78,79]. We performed the desiccation assay using osm-9 and ocr-4;ocr-2;ocr-1 (designated as ocr-1,2,4) mutant dauers. osm-9 mutants were very sensitive to desiccation although ocr-1,2,4 mutants were desiccation tolerant (Figure 5B). Thus, hygrosensation and osmosensation might involve similar molecular mechanisms.

To corroborate our results on the involvement of head neurons in hygrosensation, we investigated whether this process requires intact amphids. The amphids on the head of the worm are openings through which sensing neurons are exposed to the environment. ASH neurons extend their cilia to the amphid. Deletion of the patched-related gene daf-6 affects amphid formation such that amphid neurons cannot reach the exterior. This results in insensitivity to chemical signals, including the dauer pheromone (thereby leading to a dauer formation deficiency phenotype) [80,81]. For the desiccation experiments, we exploited the fact that daf-6 mutants can form proper dauers if cholesterol is substituted by lophenol for two generations [82]. daf-6 dauers were very sensitive to desiccation at 60% RH (Figure 5B). This result, together with the aforementioned results, underlines the importance of head neurons in hygrosensation.

Materials and MethodsWorm Cultures and Strains

C. elegans wild-type (N2), C54E4.2(gk1083);alh-2(gk3053), cdr-2(ok1996), cdr-3(ok864), cex-1(ok3163), cex-2(ok2767), ctl-1(ok1242), ctl-3(ok2042), daf-2(e1370), daf-6(e1377), dur-1(ok1010), F08H9.4&srz-97(ok1976), fat-1(ok2323), fat-3(wa22), fat-4(ok958), fat-4(wa14);fat-1(wa9), fat-5(tm420), fat-6(tm331), fat-6(tm331);fat-5(tm420), fat-6(tm331);fat-7(wa36), fat-7(wa36), fat-7(wa36);fat-5(tm420), hsp-12.6(ok1077), ocr-4(vs137);ocr-2(ak47);ocr-1(ok132), odc-1(pc13::Tc1), osm-9(ok1677), osm-11(rt142), sod-1(tm776), sod-3(gk235), sod-5(tm1146), spds-1(ok3421), and ugt-1(ok2718) strains were obtained from Caenorhabditis Genetics Center, USA, which is funded by the NIH Office of Research Infrastructure Programs (P40 OD010440). C04G2.2(tm3841), cyp-33C9(tm3809), djr-1.1(tm918), djr-1.2(tm951), F08H9.3(tm5012), hsp-70(tm2318), glod-4(tm1266), gpx-2(tm2895), gpx-6(tm2535), gpx-7(tm1990), and try-5(tm3813) strains were obtained from the National Bioresource Project, Japan. C04G2.2, cex-1, cex-2, daf-2, daf-6, djr-1.1, djr-1.2, dur-1, F08H9.4, fat-3, fat-4, fat-5, fat-6, fat-7, gpx-2, gpx-7, hsp-70, odc-1, osm-9, osm-11, sod-1, sod-5, try-5 and ugt-1 mutants were outcrossed with the wild type before or during functional analyses to eliminate the possibility that a background mutation causes the desiccation sensitivity phenotype.

Worms were maintained on NGM agar plates seeded with Escherichia coli NA22 at 15 °C [100]. Large quantities of dauers for RNA and protein extraction were obtained by growing daf-2 eggs at 25 °C in liquid culture [101]. daf-2 L3 larvae were also produced the same way, only this time by growing at 15 °C. Small quantities of dauers for desiccation assays were grown on sterol-depleted lophenol-substituted agarose plates in two generations [82] for all strains except daf-2 and daf-2;∆∆djr. Gene silencing via RNAi was performed by growing daf-2 eggs into dauers at 25 °C on E. coli HT115(DE3) expressing the dsRNA of the gene of interest [102]. These bacterial clones were purchased from Source Bioscience, UK. Their identities were confirmed by sequencing.

Desiccation Assay

Dauer larvae were washed off the plates and collected in distilled water. Dense suspensions of worms in 5 µl droplets were transferred onto 3.5 cm plastic Petri dishes in duplicates. The dishes were then placed in controlled humidity chambers equilibrated at 98% RH and preconditioned for 4 days [19]. Afterwards, one replicate of each duplicate dish for each strain was transferred into another chamber at 60% RH, whereas the other duplicate was kept in the same 98% RH chamber. After 1 day, all dishes were removed and the worms on the surface of the dishes were rehydrated with distilled water for at least 2 h. Subsequently, worms were transferred to individual 6 cm NGM agar plates seeded with E. coli NA22 and kept at 15 °C overnight to recover.

These experiments were repeated on different days for each mutant or RNAi treatment, at least twice. In each experiment, wild type or RNAi non-treated daf-2 dauer larvae were used as the positive control. The day following rehydration, survivors were counted in the total population and survival rates were calculated as the percentage of worms that survived the whole procedure. On average, 424 ± 363 (median ± median absolute deviation) worms were counted for each replicate. Differences in survival levels at 98% and 60% RH were separately compared to the relevant controls (wild type or daf-2) using beta regression [34,103]. The beta distribution of survival rate data was confirmed by one-sample Kolmogorov-Smirnov test. Mean survival rates and their standard errors were also estimated by beta regression. The desiccation sensitivity phenotype of mutants was then categorized into 4 groups: Desiccation tolerant (p ≥ 0.05 compared to the control by beta regression), desiccation sensitive (> 50% survival, p < 0.05), very sensitive (25 - 50% survival, p < 0.05) and extremely sensitive (< 25% survival, p < 0.05).

Induction of Desiccation-Related Genes and Proteins

Based on our previous results, daf-2 and wild-type dauer larvae are similar in respect to all parameters that we measured (e.g., morphology, SDS resistance, desiccation tolerance, longevity, etc.) [19,82]. In addition, daf-2 eggs grown at 25 °C in a liquid culture environment yield a large homogeneous dauer population with no prior desiccation experience. Therefore, for microarray, geLC-MS/MS, and 2D-DIGE analyses of desiccation-induced transcripts and proteins we used daf-2 dauer larvae grown in liquid culture. These worms, collected in distilled water, were first filtered on 8 µm Isopore TETP membranes (Millipore, USA) and then placed in a desiccation chamber equilibrated at 98% RH [19]. Worms for RNA extraction were kept in this chamber for 24 h to minimize RNA degradation and then collected in distilled water. Worms for protein extraction were kept in the preconditioning chamber for 4 days. Subsequently, they were collected either in SDS lysis buffer (150 mM NaCl, 50 mM Tris-HCl, 1 mM EDTA, 1% SDS (w/v), 0.2% CHAPS (w/v), 0.1% OGP (v/w), 0.7% Triton X-100 (w/v), 250 ng/ml DNase, 250 ng/ml RNase, and 1× protease inhibitor mix (Roche, Germany), pH 7.5) for geLC-MS/MS, or in urea lysis buffer (7 M urea, 2 M thiourea, 30 mM Tris, 4% CHAPS (w/v), and 1× protease inhibitor mix (GE Healthcare, Germany), pH 9.1) for 2D-DIGE. The samples were immediately frozen in liquid nitrogen until RNA and protein extraction.

One of the 4-day desiccated samples removed from the chamber was rehydrated and kept in distilled water for 24 h. These worms were then pelleted and resuspended in urea lysis buffer for subsequent 2D-DIGE analysis.

RNA and Protein Extraction

Samples for RNA extraction were homogenized by freezing in liquid nitrogen and thawing in a warm water bath without sonication. Total RNA was then extracted using TRIzol Reagent and a PureLink RNA Mini Kit (Invitrogen, Germany). The amount and purity of the RNA were determined by measuring absorbance at 230, 260, and 280 nm using a NanoDrop ND-1000 UV-VIS Spectrophotometer (Thermo Scientific, USA). RNA quality was determined on a 2100 Bioanalyzer (Agilent, Germany) using an RNA 6000 Nano Kit (Agilent, Germany).

Samples for protein extraction were homogenized by freezing in liquid nitrogen and thawing in a water bath with sonication. Samples were thawed in SDS lysis buffer at 60 °C to increase the solubilizing effect of SDS. Homogenates were then centrifuged at 20,000 g to pellet the debris. Subsequently, supernatants were transferred to Nanosep omega membrane centrifugal devices with a molecular weight cutoff of 3 KDa (Pall Corporation, USA) and concentrated by centrifuging at 15,000 g. The amounts of protein in the SDS and urea lysis buffers were quantified using a Micro BCA protein assay kit (Thermo Scientific, USA) and an RC/DC protein measurement kit (Biorad, Germany), respectively.

Microarray Hybridization and Data Analysis

4×44K oligoarrays (V2: 020186) were purchased from Agilent, Germany. Labeling and hybridization of RNA were performed as recommended by the manufacturer. Four replicates with the highest RNA quality for each of the two conditions (with or without preconditioning) were used. After hybridization, slides were scanned on an Agilent Microarray Scanner C and expression data were generated via Agilent Feature Extraction Software 10.7.1.1.

The default annotation information of Agilent V2: 020186 microarrays was outdated; therefore, it was updated with more recent annotations from WormBase release WS227 by aligning the probe sequences to the C. elegans genome and transcriptome. According to this realignment, 20,058 of 20,470 protein-coding genes of C. elegans (98%) were targeted by 34,290 probes (1.71 probes/gene).

Probe intensity data were analyzed in R statistical software using the limma package [104] as a part of the BioConductor project [105]. Raw expression data from all chips were imported as median probe intensities and variance-stabilizing normalization was applied [106]. Subsequently, control and deprecated probes were filtered out and the resulting probe-level expression data were summarized into gene-level expression values by Tukey’s median polish algorithm [107,108]. Differential expression of each gene between the two conditions was then calculated using empirical Bayes statistics. Multiple hypothesis testing with the Benjamini-Hochberg correction was applied on calculated p-values [109]. A change in the expression level was considered to be significant if the adjusted p-value was less than 0.001.

The log₂-scaled FCs (logFCs or differential expression levels) of significantly upregulated and downregulated genes were separately clustered via k-means clustering into four groups for each type of regulation: very low, low, medium, and high (Figure 1A), which were named FCCs. Genes with the lowest logFCs (FCC “very low”) were discarded because their differential expressions were not considered to be biologically significant (Figure 1A). Microarray data are available in the ArrayExpress database (www.ebi.ac.uk/arrayexpress) under accession number E-MEXP-3899.

Functional annotation clustering of the genes that showed significant changes in expression was performed using the DAVID platform v6.7 [31,32]. Functional annotation data from all available Gene Ontology and protein domain databases that are integrated into DAVID were used. Clustering was run with the following parameters: similarity term overlap = 3, similarity threshold = 0.2, initial group membership = 3, final group membership = 3, multiple linkage threshold = 0.5, and EASE = 0.05.

geLC-MS/MS

Differential expression of genes was investigated at the proteome level by geLC-MS/MS as described previously [33]. Briefly, 50 µg of total protein (as determined using the BCA assay) from biological triplicates of three samples (control, 1 day preconditioned, and 4 days preconditioned) were first separated by one-dimensional SDS-PAGE. Then, each lane was cut into five slices that were independently in-gel digested with trypsin, extracted, and vacuum-dried according to a standard protocol [110]. Dried digests were dissolved in 1% formic acid and injected into a Dionex Ultimate 3000 nano HPLC system (Dionex, Germany) directly coupled to an LTQ Orbitrap Velos (Thermo Fisher Scientific, Germany) as described previously [33]. Linear elution gradients of 150 minutes were used. Database searches were performed using Mascot 2.2.04 software (Matrix Science, UK) against the WormBase WS228 protein sequence database. Subsequently, label-free feature extraction and alignment were performed using a customized version of SuperHirn 0.03. Data normalization and peptide-level ANOVA were performed using DanteR 1.0.1.1. False discovery rates were calculated at three stages: At the level of peptide identifications, Mascot’s decoy database option was used with a 5% threshold. At the level of protein identification, the Trans-Proteomic Pipeline’s Protein Prophet tool was used with a probability cutoff of 95%. Finally, to judge the relevance of protein quantifications, the Benjamini-Hochberg FDR method was applied and an adjusted p-value cutoff level of 0.05 was used for significance.

2D-DIGE

Fifty micrograms of each protein sample (as determined using RC/DC kit) in urea lysis buffer were labeled with 250 pmol CyDye DIGE Fluor dyes (GE Healthcare, Germany). After labeling, excess dyes were quenched with 10 nmol L-lysine and the samples were reduced in rehydration buffer (7 M urea, 2 M thiourea, 4% CHAPS (w/v), and 50 mM DTT). The samples were pooled and supplemented with ampholytes (BioLytes pH 3–10, BioRad, Germany) in a total volume of 400 µl. This labeled protein mixture was loaded into an immobilized pH gradient strip (linear pH of 3–10) via passive rehydration for 24 h. Following rehydration, isoelectric focusing was performed in a Protean IEF cell (BioRad, Germany) for 55–60 kVh in total. The strip was then equilibrated in equilibration buffer (6 M urea, 2% SDS (w/v), 50 mM Tris, 20% glycerol (v/v), and 130 mM DTT) for 10 min before being placed on a 20 cm wide 12.5% SDS-polyacrylamide gel. Proteins in the strip, along with PageRuler Plus prestained protein weight marker (Fermentas, USA), were separated by SDS-PAGE at 200 V for 6 h. Finally, the gel was scanned using a Typhoon 9410 Variable Mode Imager (GE Healthcare, Germany) at 100 µm/pixel resolution for Cy2 (488 nm excitation, BP 520/40 emission filter), Cy3 (532 nm excitation, BP 580/30 emission filter), and Cy5 (633 nm excitation, BP 670/30 emission filter) at empirically determined photomultiplier tube voltages. After laser scanning, gels were stained with Comassie blue and the spots of interest were cut out. The proteins in these gel slices were extracted and characterized in a similar way as geLC-MS/MS.

Fatty Acid Analysis by LC-MS

Total lipids were isolated from preconditioned wild-type as well as fat-1, fat-3, fat-4, and fat-1,4 mutant dauer larvae via Bligh and Dyer’s method [111]. The organic phase of each sample, normalized by total protein (using BCA assay), was analyzed by reversed-phase liquid chromatography mass spectrometry using an Agilent G1312A pump equipped with an Agilent Autosampler G1329A. Separation employed an Eclipse XDB-C18 column (15 cm × 4.6 mm i.d., 5 µm, Agilent) connected to a Symmetry C18 column (7.5 cm × 4.6 mm i.d., 3.5 µm, Waters) interfaced to a Waters/Micromass LCT TOF mass spectrometer equipped with an ESI. Fatty acid species were separated by isocratic elution using 0.15% ammonium acetate in MeOH/water/MeCN (80:15:5 v/v/v) at 40 °C. The flow rate was set to 1 ml/min with a split to 30 µl/min. The mass spectrometer was operated with a spray voltage of 3 kV and a source temperature of 120 °C. Nitrogen was used as the cone and nebulizing gas. Mass spectra were acquired from the m/z range of 100–1000 controlled with Waters/Micromass MassLynx 4.1 software.

Bioinformatics Analysis of Uncharacterized Proteins

Protein sequences of uncharacterized proteins were first submitted to conserved domain searches [112] and SMART [113]. For the functional prediction of proteins, domains of unknown function were submitted to HHPRED [114] to detect remote sequence similarity to functionally annotated conserved domains. To integrate the uncharacterized proteins to potential functional networks or pathways, they were submitted to the STRING database [72]. High confidence (0.7) was chosen with all potential protein-protein interaction data sources enabled. The networks were downloaded and imported into Cytoscape [115] for further annotation.

Sequence Similarity Analysis of IDPs

Caenorhabditis elegans DUR-1C (NP_501787.3) protein sequence was compared to Adineta ricciae LEA-1A (ABU62809.1), Adineta vaga LEA-1B (ADD91471.1), Arabidopsis lyrata lyrata LEA (XP_002863597.1), Brachionus plicatilis LEA-1 (ADE05593.1), Caenorhabditis briggsae LEA-1 (XP_002637990.1), Caenorhabditis ramenei DUR-1 (XP_003089862.1), Caenorhabditis ramenei LEA-1 (XP_003116339.1), Caenorhabditis elegans DUR-1A (NP_501786.2), Caenorhabditis elegans LEA-1B (NP_001256170.1), Medicago truncatula LEA (XP_003609877.1) and Oryza sativa LEA (NP_001049087.1) protein sequences. Hordeum vulgare LEA (ABS85196.1) sequence was uaed as the out-group. The tree in Figure S3 was constructed using Geneious software with the following parameters: Blosum62 cost matrix, gap opening penalty = 12, gap extension penalty = 3, global alignment with free end gaps, Jukes-Cantor genetic distance model, and neighbor-joining method.