Foreign genes and novel hydrophilic protein genes participate in the desiccation response of the bdelloid rotifer Adineta ricciae

Certain organisms are able to survive almost complete loss of their internal water, entering a state of suspended animation called anhydrobiosis. When water becomes available again, the organism revives and resumes its normal activities. These organisms have been called ‘sleeping beauties’ since they appear not to age during dormancy (Hengherr et al., 2008; Ricci and Covino, 2005). Desiccation tolerance is relatively widespread among prokaryotes and unicellular eukaryotes [e.g. baker's yeast, Saccharomyces cerevisiae (Potts, 1994)], but is less common in multicellular eukaryotes, being exclusive to some life stages of plants (mainly seeds and pollen), a number of resurrection plants, and some invertebrates, with well-known examples among the arthropods, tardigrades, nematodes and bdelloid rotifers (Alpert, 2006; Clegg, 2001).

In some organisms, an increase in levels of di- and oligosaccharides, especially the non-reducing sugars trehalose (in animals) and sucrose (in plants), has been associated with the ability to survive desiccation (Crowe et al., 1998; Hoekstra et al., 2001). For example, the anhydrobiotic larva of the chironomid Polypedilum vanderplanki accumulates ∼18% trehalose by dry weight as it loses water, and the level of trehalose correlates with variation in survival among individuals (Watanabe et al., 2002). Such correlations, together with the powerful stabilising properties of trehalose in vitro (Colaco et al., 1992), have led to models of anhydrobiosis in which non-reducing disaccharides play a central role, and most frequently are proposed to have water replacement or vitrification functions (Crowe et al., 1998; Crowe et al., 1992). However, this is not the whole story (Tunnacliffe and Lapinski, 2003): bdelloid rotifers, for example, contain no trehalose and apparently lack the genes to synthesise it (Caprioli et al., 2004; Lapinski and Tunnacliffe, 2003). In the yeast S. cerevisiae, whose desiccation tolerance is well attested, trehalose synthesis can be abolished by mutation with only a modest reduction in survival (Ratnakumar and Tunnacliffe, 2006).

Other molecules must contribute to desiccation tolerance, therefore, and candidates have emerged from studies in several organisms. These include the hydrophilic late embryogenesis abundant (LEA) proteins, molecular chaperones, amphiphiles and antioxidants (Burnell and Tunnacliffe, 2010; Hoekstra et al., 2001; Ingram and Bartels, 1996). The latter category, for example, is likely to be important because desiccation disrupts the function of electron transfer chains, thereby elevating levels of reactive oxygen species (ROS) which have a detrimental effect on cell components (Kranner and Birtic, 2005; Rizzo et al., 2010). Direct genetic evidence of the involvement of antioxidants in anhydrobiosis was obtained recently when knockdown of glutathione peroxidase transcripts in the nematode Panagrolaimus superbus was shown to decrease survival after desiccation (Reardon et al., 2010). The LEA proteins were first discovered in plant seeds but were later found in various invertebrates, suggesting a conserved response to desiccation stress (Tunnacliffe et al., 2010). This diverse class of proteins is characterised by a versatility of function probably relating to their lack of defined structure: they are examples of intrinsically disordered proteins (IDPs) (Tompa, 2009). Despite their disordered nature, IDPs are known to carry out a range of different, and sometimes multiple, functions (Tompa et al., 2005), and LEA proteins are not unusual in this respect. Thus, LEA proteins have been implicated as molecular shields or chaperones, membrane protectants, ion sinks, hydration buffers and antioxidants (Battaglia et al., 2008; Shih et al., 2008; Tunnacliffe et al., 2010; Tunnacliffe and Wise, 2007).

Despite this progress, a more comprehensive approach to determine key adaptations in desiccation tolerance, collectively called the ‘desiccome’ (Potts, 2004), is required and gene discovery programmes are well advanced in a number of organisms. Thus, experimental approaches that identified genes upregulated by dehydration were first established in the resurrection plants (Bartels, 2005; Illing et al., 2005) and have more recently been implemented in invertebrates, primarily in tardigrades (Mali et al., 2010) and nematodes (Gal et al., 2003; Reardon et al., 2010; Tyson et al., 2007). Studies of this type could be particularly informative in bdelloid rotifers, as the anhydrobiology of these aquatic micro-invertebrates has unusual aspects: besides their lack of non-reducing disaccharides, the bdelloid LEA proteins defined to date are not especially hydrophilic and can exhibit secondary structure in solution (Pouchkina-Stantcheva et al., 2007). Furthermore, the genetics of bdelloid rotifers is also unusual: they are the only well-characterised anciently asexual metazoans, having reproduced exclusively by thelytoky (the production of females parthenogenetically) for at least 35 million years (Mark Welch and Meselson, 2000; Rice and Friberg, 2007), and they show a high frequency of horizontal gene transfer (HGT; also known as lateral gene transfer) in sub-telomeric chromosome regions (Gladyshev et al., 2008). To begin to understand the desiccome of the bdelloid rotifer, Adineta ricciae (Segers and Shiel, 2005), we produced an expressed sequence tag (EST) library enriched for genes that are active during desiccation stress and characterised a subset of clones of both recognised and novel sequences. Some sequences in the former category are apparently not of metazoan origin, suggesting that foreign genes have been appropriated for the desiccation stress response. In addition, analysis of ESTs encoding novel proteins reveals a high proportion of new classes of hydrophilic proteins other than LEA proteins, suggesting this feature as an intrinsic element of the bdelloid desiccome.

The bdelloid rotifer Adineta ricciaeSegers and Shiel, 2005 (formerly known as Adineta sp. 1) was used as a model to study gene regulation in a desiccation-tolerant organism. Clonal cultures of A. ricciae were maintained in plastic bottles or flasks; we used either plastic roller bottles (Corning Life Sciences, Amsterdam, The Netherlands) kept horizontal to maximize surface-to-volume ratio, with about 300 ml of autoclaved ultra high purity (UHP) water, or cell culture flasks (between 35 and 175 cm², with filter caps) (Nunc, Roskilde, Denmark) with 12–200 ml of the same water. Rotifers were kept at 22°C and fed once every 1–3 days with RAGO (Rotifer and Artemia GrowOut; www.aquaculturesupplies.co.uk: 15 g l^–1 stock solution prepared in water, autoclaved and allowed to cool and sediment; the supernatant was used for feeding) or Escherichia coli [grown in Luria broth (LB) medium overnight, recovered by centrifugation and resuspended in autoclaved UHP water]. Both feedstocks were stored frozen and kept at 4°C for several days during use.

A. ricciae has previously been shown to survive desiccation at rates of up to ∼80% (Ricci et al., 2004; Ricci and Covino, 2005) and similar survival rates are observed in our hands, depending on the drying protocol used. Rotifers to be desiccated were collected by filtration: the container was shaken a few times to detach rotifers, then the rotifer suspension was filtered through 20 μm or 5 μm Nitex filters (Sefar, Heiden, Switzerland). Animals collected on filters were desiccated according to one of three different drying protocols (Ricci et al., 2003). For the first two methods, an environmental test chamber (Temperature Applied Sciences Ltd., Goring, UK) was used, where temperature and relative humidity (RH) can be controlled. In the first protocol, Ricci B, RH is reduced from 98% to 40% over 15 h, then maintained at 40% RH for another 153 h; in the second protocol, Ricci C, RH is maintained at 98% RH for 72 h. In the third protocol, which was used for absolute mRNA quantification of six genes, the filter was placed between two Whatman papers soaked in 1 ml of autoclaved UHP water and left sealed at room temperature for 24 h (RH ∼100%); the Petri dish was then opened and left for 48 h until the filter had equilibrated with ambient RH (∼33% RH). Control, non-dried rotifers were also collected on filters but RNA was extracted immediately.

RNA was extracted using TRIzol reagent (Invitrogen, Paisley, UK) according to manufacturer's protocol, washed in ethanol and resuspended in diethylpyrocarbonate-treated water. RNA concentration was measured with a NanoDrop (ND-1000) spectrophotometer (Thermo Fisher Scientific, Loughborough, UK) and quality assessed by analysing the absorbance spectra and the A260/280 and A260/230 ratios.

A Super SMART™ PCR cDNA Synthesis Kit was used for synthesis of primary cDNA. A cDNA subtraction library was produced with the PCR-select cDNA subtraction kit (both from Clontech-Takara Bio Europe, Saint-Germain-en-Laye, France), following manufacturer's instructions. Briefly, rotifers were collected on filters and RNA was extracted either immediately (control sample, 0 h desiccation), or after dehydration for 24 h at 98% RH. RNA was extracted from both samples; cDNA synthesis was performed for both driver (non-desiccated) and tester (desiccated) sample; concentration was estimated by absorbance at 260 nm; and integrity was confirmed by gel electrophoresis. cDNA was amplified by long distance PCR, purified with phenol–chloroform, digested with RsaI, re-purified and diluted to a final concentration of 280 ng μl^–1. Adaptor ligation was followed by hybridisation and selective amplification of differentially expressed fragments, which were then subcloned into pCRII-TOPO vector using a TOPO TA Cloning Kit Dual Promoter (Invitrogen) and transformed into competent E. coli cells. Bacteria were grown overnight at 37°C in LB broth and 50 μg ml^–1 ampicillin, and plasmid DNA extracted with a plasmid mini- or midi-prep kit (Qiagen, Crawley, UK), and checked by restriction digestion. ESTs were sequenced by the dideoxy method at the University of Cambridge Department of Biochemistry Sequencing Facility.

DNA sequences from the subtraction library were analysed with various software packages: 4Peaks (version 1.7.2, www.mekentosj.com), ApE (A plasmid Editor, version 1.17, http://www.biology.utah.edu/jorgensen/wayned/ape) and Geneious Pro (version 4.8.4, www.geneious.com). Failed sequences (e.g. sequences with indistinguishable peaks, or with multiple/overlapping readings) were discarded at this stage and approved sequences were analysed with blastx (Altschul et al., 1990) (blastx versions 2.2.21–2.2.23, non-redundant protein database). Further analysis on selected ESTs was carried out with tblastx.

Sequences with an E-value lower than E–05 in the blastx search were considered to be defined hits whereas sequences with an E-value higher than E–05 were considered novel. In the first case, sequences were assigned to one of 10 functional categories. Where no match was found with blastx, the six possible open reading frames (ORFs) were obtained with Geneious Pro and the longest was further analysed. Physico-chemical characteristics were inferred with different software packages. The presence of a spliced leader, found at the 5′ end of a proportion of bdelloid mRNAs (Pouchkina-Stantcheva and Tunnacliffe, 2005), and poly(A) tail sequences was determined with Geneious; hydrophilicity was evaluated using ExPASy ProtScale (http://www.expasy.ch/cgi-bin/protscale.pl); theoretical pI, number of charged residues, GRAVY index (grand average hydropathy) and glycine content were calculated with the ExPASy ProtParam tool (http://www.expasy.ch/tools/protparam.html); disorder predictions were performed at PONDR (Predictors of Natural Disordered Regions; http://www.pondr.com) using Uversky plots, VL-TX and VL3 algorithms (Li et al., 1999; Radivojac et al., 2003), and with FoldIndex (Prilusky et al., 2005) (http://bip.weizmann.ac.il/fldbin/findex).

To support the foreign origin of some ESTs, phylogenetic analyses were performed using Geneious. ORFs were analysed with blastp and amino acid sequences corresponding to the ten best species matches were downloaded and aligned with ClustalW, and a Neighbour-Joining tree was built (Jukes-Cantor genetic distance model; without designating any outgroup a priori). Bootstrap support was calculated with 1000 iterations, and taxa were colour-coded in the consensus tree. Sequences were eliminated where long branches caused all other branches to collapse.

Gene expression profile before and after dehydration was usually studied by relative real-time quantitative PCR (qPCR), normalising gene expression against a reference gene whose expression level was expected to remain constant, or nearly so (either actin or 18S or both). Template cDNA was obtained as described above and all PCR run in a RotorGene 3000 using the SYBR-Green PCR kit (Qiagen). In a number of cases, absolute quantification was performed with the same instrument using the Real-Time RT-qPCR kit (Qiagen), as described previously (Browne et al., 2004). In this case, plasmid DNA from bacterial stocks was extracted with a midi-prep kit (Qiagen), quantified by ultraviolet absorbance with a NanoDrop 1000 spectrophotometer (Thermo Fisher Scientific), in vitro transcribed with a MEGAscript Kit (Applied Biosystems/Ambion, Warrington, UK) according to manufacturer's instructions using either T7 or SP6 polymerase depending on fragment orientation, and the total number of target molecules quantified by serial dilution (generally in the range of 0.001–1000 ng μl^–1).

Bdelloid rotifers require a slow rate of drying for maximum survival (Lapinski and Tunnacliffe, 2003; Ricci et al., 2003), and we therefore reasoned that a number of metabolic adaptations need to be activated for successful anhydrobiosis in these invertebrates. To attempt to identify adaptations regulated at the transcriptional level, we generated an EST library enriched for sequences that are over-represented in the bdelloid transcriptome after 24 h desiccation compared to non-dried controls. The initial 93 readable sequences were further analysed for possible duplicates, returning a total of 75 unique putative candidates that were then compared with known sequences using the BLAST search tool. Blastx returned 36 sequences with a significant match in the databases (listed in Table 1), and implicated the remaining 39 ESTs as novel sequences (nearest match scoring >E–05). Of these, eight ESTs (F24-15, F24-19, F24-26, F24-31, F24-38, F24-54, F24-91 and F24-98; accession numbers HO188837, HO188839, HO188841, HO188843, HO188845, HO188853, HO188864 and HO188866, respectively) did not show any clear ORFs (≥50 amino acids), and were therefore excluded from further analyses; the remaining 31 ESTs are listed in Table 2. Three of the sequences without clear ORFs (F24-15, F24-38 and F24-98) were included in expression profile studies (see below).

The 36 sequences with a significant match in protein databases can be placed in various categories as shown in Table 1. The largest two categories are metabolism and stress, consistent with the need to reprogramme cell biochemistry and to launch damage prevention and repair systems in the face of desiccation stress. The former category is diverse, although several examples relate to carbohydrate metabolism. In the latter category, more than half the ESTs (F24-6, F24-30, F24-61b, F24-74) represent genes concerned with oxidative stress, in line with the importance of redox balancing and ROS management during desiccation. Two further sequences are implicated in DNA repair, an essential function after the DNA damage resulting from desiccation (Mattimore and Battista, 1996). F24-37 identifies an E2 ubiquitin-conjugating enzyme (RAD6) required for lesion repair, and F24-119 represents a DNA polymerase λ, involved in double strand break repair. An LEA protein sequence is also found in the stress category, and this corresponds to one of the pair of very similar LEA protein genes previously described for A. ricciae (Pouchkina-Stantcheva et al., 2007). Other categories with small numbers of ESTs are protein homeostasis, cell structure, RNA processing, apoptosis, ribosomal proteins, signalling, transport and translation (Table 1). In the cell structure category, four actin ESTs were found with a few differences in the nucleotide and amino acid sequences (supplementary material Fig. S1); where actin was used as a reference gene in quantitative PCR experiments, the sequence corresponding to F24-59 was used.

Included in Table 1 for each bdelloid EST is the species giving the best blastx match against the set of known protein sequences. Intriguingly, six of these, shown by grey shading in Table 1, gave closest matches to non-metazoan species. Two more sequences, F24-11 and F24-86, also gave a non-metazoan species as a top hit, but the E-values were relatively high and therefore these ESTs were not analysed further. Since bdelloid rotifers have been shown recently to accumulate foreign DNA sequences from bacterial, fungal and plant sources on a large scale, at least some of which are known to be expressed (Gladyshev et al., 2008), an additional test was performed to examine the evolutionary origins of these six sequences. A second BLAST algorithm, tblastx, which translates each EST and compares potential ORFs against a database of translated non-redundant nucleotide sequences, was used, and separate lists were compiled of best hits against targets among the metazoa, plants, fungi, ‘other’ (anything excluding metazoans, plants, fungi, eubacteria, viruses and Archea) and eubacteria (Table 3). Two of the six ESTs were found to give matches against metazoans: F24-74 returned an E value of 4E–11 for a predicted protein from Saccoglossus kowalevskii, a hemichordate, and a second hit for Drosophila sechellia (1E–07), while F24-8 returned an E-value of 3E–16 against a sequence from a hypothetical protein from Branchiostoma floridae, a cephalochordate, a second hit for the stony coral Acropora millepora, via a Total Shotgun Approach (4E–16), and a subsequent hit for the human body louse (1E–15). Therefore, although the ESTs F24-74 and F24-8 gave considerably lower tblastx E-values against non-metazoans, there was sufficient doubt to eliminate them from further consideration.

For the remaining four ESTs, however, the tblastx results supported an exogenous origin for the corresponding genes. F24-2, whose sequence shows similarity to glucose-repressible genes of unknown function, gave no hits with an E-value <1 among the metazoa, but confirmed Podospora anserina as the best match (4E–14) among the fungi. Hits were also obtained among the plants, with Oryza rufipogon at 6E–09, but no bacterial or protist sequences with significant matches were returned. F24-3, which probably represents a phosphate transporter gene, gave a top tblastx match of 2E–43 against a sequence from Ustilago maydis, a fungal plant pathogen, whereas the best metazoan example was a predicted sequence from honey bee (Apis mellifera), which returned a not-significant E-value of 0.008. Blastx for F24-16 identified a dioxygenase gene, scoring best against an alpha proteobacterium, and tblastx also gave the most significant hit against a bacterium, and a slightly less significant hit for a green alga and a protist, without returning any significant hit among metazoans. Finally, in Table 1, the best blastx match for F24-20 was with Methylobacterium chloromethanicum CM4; the tblastx algorithm also returned a Methylobacterium species, albeit different, as the top hit with an E-value of 2E–32. The best-matching metazoan sequence was from human with an E-value of just 0.87, and a fungal sequence gave a marginally significant hit (3E–04). To further confirm the non-metazoan nature of these four sequences, phylogenetic analyses were carried out, confirming a bacterial origin for F24-20 (Fig. 1) and F24-16 (supplementary material Fig. S2A), and a fungal origin for F24-3 (supplementary material Fig. S2B). F24-2 did not produce any matches with metazoa (supplementary material Fig. S2C). In summary, therefore, at least four ESTs in our panel seem to represent genes in the bdelloid rotifer genome that derive from horizontal gene transfer.

For the 31 of the 39 novel sequences in the EST library with an identifiable ORF, we used bioinformatics tools to provide insight into some of the potential physico-chemical properties of their predicted proteins. In particular, we were interested in those proteins that might play an analogous role to LEA proteins in other species, since the bdelloid LEA proteins characterised to date seem atypical: they are not particularly hydrophilic, and one of them has secondary structure in the hydrated state (Pouchkina-Stantcheva et al., 2007) (Table 2). We therefore looked for examples of predicted proteins which were both hydrophilic and potentially disordered, assessing the former property using the GRAVY index and the latter with three different algorithms, VL-TX, VL3 and FoldIndex, and a Uversky plot of mean net hydropathy against mean net charge (Table 2). We also calculated the theoretical isoelectric point (pI) for each protein, the percentage of charged amino acids and the glycine content. For comparison, the same analysis was performed for BSA (bovine serum albumin), as a typical soluble globular protein; for the two previously-reported LEA proteins from A. ricciae (ArLEA1A and ArLEA1B) (Pouchkina-Stantcheva et al., 2007); and for two more typical hydrophilic IDPs from the anhydrobiotic nematode Aphelenchus avenae: AavLEA1, a group 3 LEA protein (Browne et al., 2002; Goyal et al., 2003), and anhydrin, a basic IDP (Browne et al., 2004; Chakrabortee et al., 2010).

As criteria for identifying hydrophilic IDPs, we asked for a GRAVY score more negative than that of BSA (–0.43), and two out of four of the following: VL-TX and VL3 indices higher than 0.5 (i.e. at least 50% disordered), a FoldIndex score less than zero (indicative of proteins that are likely to be unfolded), and location in disordered space in the Uversky plot. Eight examples (grey shading in Table 2) met these criteria; of these, two (F24-47, F24-66) are acidic (pI<6.0), one is broadly neutral (F24-106), and five (F24-12, F24-39, F24-49, F24-82, F24-104) are basic (pI>9.0). Thus, eight out of 31, or 26%, novel sequences are likely to be hydrophilic IDPs, or at least to contain intrinsically disordered regions (IDRs). With a more stringent approach, requiring three out of the four criteria for disorder to be met, three ESTs (F24-49, F24-82, F24-106) were discarded (light grey shading in Table 2), leaving five out of 31, or 16%, as hydrophilic IDPs. Of the EST dataset as a whole (67 ESTs in total, excluding the eight ESTs without a recognisable ORF), hydrophilic IDPs constituted 12% or 7%, respectively, depending on the selection criteria used. This is within the range of values for IDP representation in eukaryote proteomes (Dunker et al., 2000; Tompa, 2009) and therefore suggests that such proteins are not over-represented in the bdelloid rotifer. The IDPs identified do not contain a high proportion of glycine and therefore do not conform to the definition of ‘hydrophilins’ (Garay-Arroyo et al., 2000), where a minimum of 6% glycine was stipulated.

To confirm the enrichment for dehydration-regulated genes in the EST library, the expression levels of 65 genes were investigated under various drying regimes using qPCR. Three different drying protocols were used, as described in Materials and methods. The results are summarised graphically in Fig. 2 and showed that at least a third of the genes, represented by 21 ESTs, are consistently upregulated, i.e. have more than a twofold increase in mRNA levels, in all the drying experiments. Of these, in some cases the upregulation was independent of the desiccation protocol (e.g. F24-2/grg1, F24-22/ArLEA1A), whereas in other cases a consistently higher expression level was observed when rotifers were subjected to protocol Ricci B (e.g. F24-36/novel, F24-50/ubiquitin-like protein). The latter observation might reflect the more stringent conditions of water loss imposed by this protocol. Another 12 genes showed inconsistent regulation, but were sometimes upregulated by dehydration (e.g. F24-4/novel, F24-45/phosphatidylinositol-4-phosphate 5-kinase). Twenty-six genes were not regulated by water loss, with a fold change in mRNA of between 0.5 and 2 (e.g. F24-11/amidase), and six genes were apparently consistently downregulated (e.g. F24-52b/novel, F24-113/RNA recognition motif). Of particular interest were the four foreign genes identified (see Tables 1 and 2). Most participated to some degree in the response to water loss: one (F24-2/grg1) was consistently upregulated by desiccation in all the experiments performed, two (F24-3/inorganic phosphate transporter and F24-20/putative major royal jelly protein) showed a variable pattern of regulation, and one (F24-16/amidase) was marginally regulated. Similarly, of the hydrophilic IDPs identified (Table 3), the majority (three out of five, using the most stringent selection criteria; see above) were upregulated by evaporative water loss, consistent with a role in desiccation tolerance. In general, 37.5% of the novel ESTs were upregulated and 12.5% were downregulated upon desiccation, while 12.5% were inconsistently regulated.

This study demonstrates for the first time that an anhydrobiotic bdelloid rotifer, A. ricciae, mounts a significant transcriptional response to evaporative water loss. It therefore seems likely that bdelloids need to make at least a partial adjustment to their biochemistry and physiology through gene regulation to survive desiccation. This is not necessarily expected, because some organisms are apparently capable of anhydrobiosis without the need for metabolic modification (Shannon et al., 2005). Some clues as to the nature of the adjustments occurring in bdelloids were obtained from examination of an EST library enriched for dehydration-regulated gene sequences. Using a subtractive hybridization approach, we identified 75 ESTs that are expressed 24 h after the onset of desiccation, of which 44% (21 defined, 12 novel) are consistently upregulated compared with the fully hydrated animal. A number of other ESTs (12 defined, four novel) represent genes that are upregulated by dehydration under some conditions, giving a total of 65% of the sequences identified that can respond positively to evaporative water loss. Approximately half of all ESTs were identifiable by database matching and their respective genes could be assigned to various broad functional categories, while the remainder had no obvious counterparts in other species and therefore represent novel sequences. Of the recognisable sequences, those in the stress and protein homeostasis categories are likely to be the most relevant to anhydrobiosis, including LEA protein, antioxidant, DNA repair, molecular chaperone and protein clearance genes. Therefore, although we are clearly only sampling a subset of the desiccome, bdelloids follow a pattern of gene expression observed in other dehydrating anhydrobiotes, principally nematodes (Adhikari et al., 2009; Gal et al., 2003; Reardon et al., 2010; Tyson et al., 2007) and tardigrades (Mali et al., 2010; Schill et al., 2004). Nevertheless, evidence for the involvement of non-reducing disaccharides in bdelloid rotifers is still lacking.

Only a single LEA protein EST was found in our dataset, corresponding to Ar-lea-1A, one of two very similar LEA protein genes described previously in A. ricciae (Pouchkina-Stantcheva et al., 2007). Both bdelloid LEA proteins are unusual compared with group 3 LEA proteins reported in other species because they do not exhibit a high degree of disorder; indeed, one of the bdelloid examples, ArLEA1B, adopts some secondary structure in solution, and can be considered a molten globule rather than an IDP proper. Moreover, N-terminal and C-terminal signal sequences suggest that both bdelloid LEA proteins are located in the lumen of the endoplasmic reticulum; preliminary experimental data are consistent with this (C. B. Tripathi and A.T., unpublished). This subcellular localisation for group 3 LEA proteins is unusual, although not unprecedented (Tunnacliffe and Wise, 2007), with most examples being found in the cytoplasm. It might be expected, therefore, that other LEA proteins with different properties are present in A. ricciae, particularly since the bdelloid rotifers' relatives, monogonont rotifers, have been shown to possess at least two different LEA proteins, both of which are more typically hydrophilic and probably located in the cytoplasm (Denekamp et al., 2010). Other invertebrates are known to contain several LEA proteins including group 1 sequences (Mali et al., 2010; Sharon et al., 2009), a class of LEA proteins previously thought to be plant specific. A more detailed analysis of the bdelloid LEA proteome will be possible once the A. ricciae genome sequence is available (C.B., A. Carr, A.T. and G. Micklem, unpublished).

The presence of only one type of LEA protein sequence in the EST dataset led us to search for other types of unstructured hydrophilic proteins that might have an analogous role in desiccation tolerance. We did not find an example of a ‘hydrophilin’ (Garay-Arroyo et al., 2000), i.e. with both high hydrophilicity and high glycine content, but we did uncover candidate hydrophilic IDPs that shared some physico-chemical characteristics with LEA proteins. Of the five strongest candidates, three are upregulated during evaporative water loss and thus participate in the desiccation stress response. In addition, at least 12 other novel sequences are also consistently upregulated by dehydration; as the function of these genes is completely unknown, they represent an interesting opportunity for further studies.

Intriguingly, at least four ESTs, i.e. 11% of the recognisable sequences, are strong candidates for horizontal gene transfer from fungal (F24-3) or bacterial (F24-2, F24-16, F24-20) origins. Moreover, because at least three of the four examples can be upregulated in response to water loss, they are not only capable of being expressed in the bdelloid rotifer, but also form part of the A. ricciae desiccome. These ESTs (and indeed all those reported in this paper) are also present in the A. ricciae genome sequence and are therefore are not the result of contamination of bdelloid cultures (C.B., A. Carr, G. Micklem and A.T., unpublished). This confirms and extends previous findings (Gladyshev et al., 2008), which demonstrated a remarkably high frequency of foreign genes in the subtelomeric regions of A. vaga and Philodina roseola chromosomes. Bdelloid rotifers, therefore, seem to be unusual among higher eukaryotes for the extent to which they assimilate foreign genes. For example, where animal genomes have been analysed, only a few cases of individual HGT are known (Keeling and Palmer, 2008). If the whole transcriptome reflects the results described in this study, bdelloids might express hundreds of foreign genes. Presumably this is adaptive, conferring some selective advantage and perhaps partially compensating for the inability of bdelloids, as asexuals, to accumulate advantageous genetic variation through sexual exchange. Gladyshev et al. have speculated that the facility to incorporate foreign DNA into their genomes results from the desiccation tolerance of bdelloids, as the dry state is expected to result in chromosome breakage and cell membrane permeability, both of which are likely to stimulate HGT (Gladyshev et al., 2008). This would rather neatly intertwine the two most remarkable characteristics of bdelloid rotifers, namely anhydrobiosis and asexual reproduction.

We thank Prof. Tim Barraclough (Imperial College, London) and Dr Esther Lubzens (Israel Oceanographic and Limnological Research) for helpful comments on the manuscript.

 
• BLAST

  Basic Local Alignment Search Tool

 
• EST

  expressed sequence tag

 
• GRAVY

  grand average hydropathy

 
• HGT

  horizontal gene transfer

 
• IDP

  intrinsically disordered protein

 
• LEA

  late embryogenesis abundant

 
• ORF

  open reading frame

 
• PCR

  polymerase chain reaction

 
• PONDR

  predictor of natural disordered regions

 
• RH

  relative humidity

 
• ROS

  reactive oxygen species