Surviving extreme polar winters by desiccation: clues from Arctic springtail (Onychiurus arcticus) EST libraries

CDNA libraries construction and characterisation

Five libraries were produced from both desiccating and re-hydrating populations of O. arcticus from stages that had previously been defined as potentially informative for molecular analyses [13].

• Library C: Controls

• Library D1: Desiccating: animals at -2°C. This is the critical temperature at which trehalose is significantly up-regulated at the expense of stores of glycogen.

• Library D2: Fully desiccated animals.

• Library R1: Animals that had been recovering for 8 hours.

• Library R2: Fully recovered animals.

The sequences from each library were further processed using Blast2GO [23], but showed very similar compositions when defined in terms of their molecular function (GO annotation level 2). An example of this output is shown for all libraries in Figure 1, with the majority of clones having either catalytic or binding activities. Although the libraries were normalised, GOSSIP [24] was used to perform statistical analyses using pair-wise comparisons of each library with the control library to identify any potential enrichment for particular genes or gene functions. No enrichment was found using the corrected p-values of the False Discovery Rate, although there were a number of significant single test p-values. For example GO016860: intramolecular oxidoreductase activity and GO008237: metalloexopeptidase activity were both elevated in the desiccated library compared to the control with single test p-values of 0.05, these were not significant using the False Discovery Rate. The problem with such pair wise comparisons is that even though the comparative library was always the control animals, the GO categories listed for each pair wise comparison varied considerably and it was not possible using this technique to make global statements of certain molecular functions being statistically enhanced in one library compared to the others. These data do indicate that the normalisation procedure in the library production process was relatively effective.

GO annotation

The decision was made to analyse the libraries using GO annotation rather than keyword extraction from Blast comment lines, as GO categories are more generic than gene names. This facilitates the comprehensive identification of gene sets associated with biochemical pathways without in-depth knowledge of individual pathway components and also where potentially multiple pathways exist for a particular process (cf. trehalose [25]). In the following discussion PGO refers to a Biological Process GO annotation; FGO is Molecular Function and CGO is Cellular component.

Initially the focus of the searches concentrated on general processes such as response to water, water deprivation, abiotic stimulus, desiccation and Qdrought recovery. Disappointingly only 11 clones were identified from the five libraries, but even so, these did include some clones of further interest. PGO:0009414, response to water deprivation identified a putative aquaporin (the same clones were also identified under a specific GO search for aquaporin) and PGO:0009415, response to water, produced ESTs with matches to dehydrin which is induced in response to water stress in plants, the latter of which would not have been identified using extraction of Blast keyword data. Given the relative lack of success of the generic searches, specific genes and pathways were targeted. Of the genes present in the libraries, there was particular interest in identifying the following genes:

• The trehalose and glycogen pathways (as trehalose is produced from glycogen during the desiccation process) [25].

• Aquaporins: 6 transmembrane domain proteins involved in water transport [26, 27].

• Genes involved in cell protection, such as antioxidants

• Genes involved in moulting, as moulting has previously been implicated in lowering the supercooling point and survival ability of Antarctic springtails (Cryptopygus antarcticus) [28, 29].

• LEA (Late Embryogeneis Abundant) proteins, which have been shown to be involved in desiccation in a number of organisms [30].

Trehalose and glycogen pathways

Aquaporins

These proteins are associated with water transport across membranes (Kruse et al, 2006) and have the GO annotation: FGO:0015250: water channel activity. Searches of all libraries revealed 7 clones in total (4 singletons and a cluster of 3) matching three potentially different aquaporin genes:

• Q0IG28: Aquaporin 1 from Aedes aegypti. Two clones from control and desiccating libraries (C and D1), matches in excess of P value = 177, expect value = 9.7e-12.

• Q9NHW7: Aquaporin AQPAe.a from Aedes aegypti. Two clones from desiccated and recovering libraries (D2 and R2), matches in excess of P value = 512, expect value = 3.2 e-47

• UniRef100 000051A00B: Predicted similar to Drip CG9023_PB isoform B Apis mellifera. 1 cluster of 3 genes from the desiccated library (D2) with a match of P value = 480, expect value = 7.7 e-44

Translation of the individual clones and alignments confirmed that three different aquaporin genes had been cloned (Figure 2A) with a maximum of 49.4% amino acid identity between clones when compared over identical lengths (Figure 2B). The clones were all around 200 amino acids in length, which is approximately 80% of the expected length of an aquaporin gene. All clones included 5 complete transmembrane (TM) domains with TM6 present in sb_006_05H07 and partial in the two other clones. All three contained the classical footprint of aquaporins: two NPA motifs, cysteine 181 was not conserved and only CL138 had the consensus site for N-linked glycosylation. The consensus site for phosphorylation by protein kinase C was not present in any of the clones [31]. Homology between the clones was relatively low at a maximum of 49.4%, but not surprising as homology within aquaporins is generally low. For example the maximum homology between the latest aquaporin identified in mouse (AQP12) with the other mouse AQPs is 38.9% [32]. When the 328 bp sequence from exon 1 of the gene for aquaporin-2 was compared in 12 mammalian species, only 14 out of 109 amino acids were conserved throughout the mammalian aquaporin family, of these 13 were conserved in the springtail AQPs (Figure 2A). Identity of the springtail clones was confirmed by phylogenetic analysis (Figure 3). All three clones clustered with the other insect aquaporins extracted from the database. Interestingly all insect clones clustered more strongly with the aquaglyceroporins, AQP8, the putative ancestral molecule [27] and AQP 11 and 12. Relatively little is known about the latter two genes, but the human AQP12 is hypothesised to play a novel intracellular role in digestive enzyme secretion [32]. Limited functional analysis has been carried out on the insect aquaporins, but evidence points toward them being designated as classical water channel molecules [33, 34]. Cicadella viridis AQPcic has been shown to increase osmotic membrane water permeability when expressed in Xenopus oocytes and the Aedes aegypti AQP (Accession number: Q9NHW7) gene was localised to tracheolar cells associated with malpigian tubules and therefore it was hypothesised that this protein played a role in the removal of tracheolar fluid during respiration.

Cell protection

Moulting

Processes involved in the survival of insects at low temperatures include the removal or deactivation of ice nucleating agents, accumulation of cryoprotectants and thermal hysteresis proteins [35–38]. Moulting has recently been shown to be associated with reduction of the supercooling point (SCP) and hence cryoprotection in Antarctic springtails (Crypotpygus antarcticus) [28, 29]. This might be expected to depress the SCP, because in Collembola the mid-gut and its entire contents are shed during moulting [39] resulting in the expulsion of potential ice nucleators in the animal gut. The physiology and timing of moulting in O. arcticu s has not been documented to date, although with a rigid exoskeleton they clearly moult regularly in order to grow. So genes and pathways involved in moulting were also investigated using GO annotations (data not shown). A number of relevant genes were identified for both juvenile hormone (a pleiotropic hormone, which in concert with ecdysteroids orchestrates moulting and metamorphosis and may be involved in reproduction in some species) and members of the ecdysone pathway. The latter included the ecdysone receptor and the protein ultraspiracle (XR2C) chorion factor. In addition to a considerable number of putative transcription factors and chromatin remodelling subunits were identified. It was interesting to note that more matches were found in the actively desiccating (D1) and recovering animals (LIBs R1 and R2) (10, 24 and 14 respectively, total = 48) compared to only 8 in the control and 5 in the desiccated populations. Thus indicating that moulting may either play a role in the desiccation process, or potentially is triggered in some animals by the cellular stress involved in desiccation and recovery. We are currently investigating the role of moulting in desiccation of O. arcticus using a biochemical test for 20-hydroxy ecdysone and are also actively collecting animals that have either just moulted or are in the process of moulting for more detailed molecular analysis.

Lea proteins

Of the genes and pathways under investigation, only LEA proteins were without GO annotations. So a search was made of the ESTs using Blast annotations. Identification of LEAs is problematical as these proteins are not highly conserved [30]. Comparison of LEA proteins between different species reveals only 53.5% identity between two cereal crops (Q42376: LEA3, maize and Q03968: LEA3, wheat), which reduces dramatically to 27.2% between different phyla such as chick pea and nematode (O49816 and Q95V77 respectively). Even the 11 amino acid repeat unit, which is a feature of LEAs shows little conservation with only 1 or 2 identical amino acids between chick pea and the nematode Aphelenus avenae. Searches of the O. arcticus BLAST annotations revealed a cluster of two clones present in Library D2 (desiccated animals) and a single clone in Library D1 (desiccating animals). This is exactly where such genes would be expected if they were involved in the desiccation process. Translation of all three clones and alignments indicated that only one gene was present. The primary BLAST match results for this clone against the databases were to Q1DH19, a putative uncharacterised protein from the yellow fever mosquito. After a number of subsequent matches to uncharacterised insect proteins, there were then matches to abhydrolase genes and LEA proteins (Oryza sativa) with probability and expect scores for the latter of 179 and 6.4e^-11 (Figure 4). At the amino acid level, identities were low with the best match to the uncharacterised protein showing 45% identity and the LEA protein, 31%.

Database searches of LEA proteins can produce matches to abhydrolase genes (as happened with this clone), due to the presence of an abhydrolase domain (denoted in Figure 4). Abhydrolases are a largely uncharacterised protein family, but members of this family do contain domains with hydrolase activity and therefore could potentially be involved in desiccation biochemistry. Phylogenetic analysis of the O. arcticus translated gene fragment suggests that this fragment most closely matches the ABHD_A (or ABHD_10) abhydrolase domain (data not shown). The question remains, is this clone sb_009_02E03 an LEA or an abhydrolase? This is virtually impossible to answer with short sequence fragments and the answer may become apparent with the full-length sequence, the cloning of which by RACE PCR is now underway. Also Western blotting studies are being carried out on O. arcticus using heterologous probing of a LEA antibody to identify such proteins via an alternative route.

Differences between the libraries

Sample collection and preparation

Onychiurus arcticus, were collected under the bird cliffs at Stuphallet and Krykkefjellet on the Brøggerhalvøya, near Ny Ålesund, Spitsbergen, Svalbard, Norway (78°56'N, 11°53'E) and transported to the British Antarctic Survey (BAS), Cambridge, for analysis. Animals (mixture of both adult and juveniles) were cultured in ventilated plastic boxes containing moss, lichen and soil taken from field sites and fed on dried baker's yeast. Cultures were kept moist at +4°C.

Five groups of animals were prepared for library production:

• Library C: Controls: live animals which were kept in a +4°C cabinet

• Library D1: Desiccating: animals were cooled for two weeks from +2 to -2°C in culture pots containing a base of wet plaster of Paris/charcoal at a rate of 2°C per week. -2°C is the critical temperature at which trehalose is significantly up-regulated at the expense of stores of glycerol.

• Library D2: Fully Desiccated: animals were cooled from +2 to -14°C in culture pots containing a base of wet plaster of Paris/charcoal at a rate of 2°C per week.

• Library R1: Recovering: animals from the -14°C group were allowed to recover at +5°C with moisture for 8 hours.

• Library R2: Fully recovered: animals from the -14°C group were allowed to recover at +5°C moisture for 24 hours.

All groups of animals were rapidly frozen in liquid nitrogen and kept at -80°C until required.

Library construction

Total RNA was extracted from frozen (-80°C) samples of O. arcticus using TRI Reagent (Sigma) according to manufacturer's instructions. cDNA was synthesized and normalized using a combination of the Trimmer-Direct kit (Evrogen, Moscow, Russia) and the SMART cDNA Library Construction kit (BD Biosciences Clontech, Palo-Alto, CA) according to the protocol from Evrogen [40]. Modifications to the protocol were made concerning the columns used for size selection and the cloning vector: to improve clone size selection Chroma spin 1000 were used instead of Chroma spin 400 (both BD Biosciences Clontech, Palo-Alto, CA) and pal32 (Evrogen, Moscow, Russia) was used for directional cloning with insertion between two SfiI sites (GGCCATTACGGCCGGG del(CATGTC) GGCCGCCTCGGCC. This procedure was chosen because of the low amount of starting material [41] and the normalisation process increased the efficiency of rare transcript discovery [42, 43]. Plasmids were transferred via electroporation to Escherichia coli (strain DH10B, Invitrogen, Karlsruhe, Germany).

cDNA Sequencing

Plasmids were isolated according to the method of [44] and 5'end sequenced using Dye Terminator Chemistry version 3.1 (ABI, Weiterstadt, Germany) and 3730XL ABI capillary sequencer systems (ABI, Weiterstadt, Germany).

Sequence analysis and EST clustering

Sequence fasta files were processed using the script Trace2dbest [45], which incorporated the phred [46, 47] and crossmatch (P. Green, unpublished) programmes. A minimum cut-off value of 150 bp was applied after quality control processing for sequence database searching and for generating the submission file for dbEST [48] (Accession numbers, dbEST: 49109381–49125759, Genbank: EW744731–EW761109). Tgicl [49] was used for clustering the fasta files, incorporating quality scores, for each of the five libraries, as well as for all the libraries together. The clusters were database searched using Blastx [50] against the Uniprot/Swissprot and Uniprot/Trembl databases [51] (at 12/06/2007), with matches annotated for all scores with an expect score in excess of 1e-10. Sequences with a database match were then further annotated using GO [52] (at 24/07/2007). Another view of the data was generated by Blast2GO [23] using the non-redundant (nr) database [53]. With the Blast2Go annotation, the programme GOSSIP [24] was run to identify any enrichment for GO annotations between the libraries. Sequence manipulation was carried out using the EMBOSS suite of programmes [54]. Sequences were clustered using ClustalW [55] and then subjected to phylogenetic analysis using the Phylip suite of programmes with bootstrapping [56] and displayed using MEGA4 [57]. Sequence alignments were displayed using BoxShade v3.21 [58].